working around the clock - repub.eur.nl · working around the clock: adverse health effects of...

Working around the Clock:Adverse health effects of circadian rhythm disturbance

Werken rond de klok:Nadelige gezondheidseffecten door verstoring van het circadiane ritme

Kirsten Van Dycke

40036 Dycke, Kirsten.indd 1 11-04-16 11:02

ISBN 978-94-6332-012-2

Design Cover: Lysette Hartman

Design Inside: Ferdinand van Nispen tot Pannerden,

Citroenvlinder DTP&Vormgeving, my-thesis.nl

Printed by: GVO Drukkers en vormgevers, Ede, The Netherlands

Copyright © 2016 Kirsten Van Dycke

All rights reserved. No part of this thesis may be reproduced, stored in a retrieval system or

transmitted in any form or by any means without prior permission of the author.


Working around the Clock:Adverse health effects of circadian rhythm disturbance

Werken rond de klok:Nadelige gezondheidseffecten door verstoring van het circadiane ritme

Proefschriftter verkrijging van de graad van doctor aan de

Erasmus Universiteit Rotterdam

op gezag van de rector magnificus

Prof.dr. H.A.P. Pols

en volgens besluit van het College van Promoties.

De openbare verdediging zal plaatsvinden op

dinsdag 7 juni 2016 om 15:30

door

Kirsten Catharina Gabriëlla Van Dycke

geboren te Arnhem


Promotiecommissie

Promotoren: Prof.dr. G.T.J. van der Horst

Prof.dr. H. van Steeg

Overige leden: Prof.dr. T. Roenneberg

Dr. E.F.C. van Rossum

Dr.ing. R.W.F. de Bruin

Copromotor: Dr. W. Rodenburg


Contents

Contents 6

Chapter 1 General introduction 9

Chapter 2 Chronically alternating light cycles increase breast

cancer risk in mice

23

Chapter 3 Biomarker discovery using a comparative

omics approach in a mouse model developing

heterogeneous mammary cancer subtypes

39

Chapter 4 Biomarkers for circadian rhythm disturbance

independent of time of day

59

Chapter 5 Diurnal variation of hormonal and lipid biomarkers in

a molecular epidemiology-like setting

77

Chapter 6 Attenuation of circadian rhythmicity in hepatic gene

expression upon chronic alternating light cycle

exposure

97


Summary and general discussion 113

Nederlandse Samenvatting 129

Curriculum Vitae 134

List of publications 136

PhD portfolio 137

Dankwoord 139

References 142

Appendix 1 A day and night difference in the response of

the hepatic transcriptome to cyclophosphamide

treatment

157

Appendix 2 Supplemental data 181


CHAPTER 1

General introduction


Chapter 1

10

The circadian clockTo anticipate the recurring environmental changes resulting from the Earth’s

rotation, most organisms have developed a circadian clock with periodicity of

approximately (circa) one day (diem). Circadian rhythms can be found in behavior,

physiology and metabolism. Well known examples are the sleep-wake cycle or

rhythm in locomotor activity (Hurd et al., 1998, Cahill et al., 1998), daily variations

in body temperature (Refinetti and Menaker, 1992) and blood pressure (Pickering,

1990) and melatonin levels that rise at sunset and decrease at sunrise (Hardeland

et al., 2006). In mammals, this internal timekeeping system is composed of a master

or central clock, the suprachiasmatic nucleus (SCN) located in the brain, and

peripheral clock in all other tissues in the body. The SCN consists of two bilateral

nuclei in the ventral hypothalamus, just above the optic chiasm and is required to

synchronize the peripheral clocks (Moore and Eichler, 1972, Moore, 1983, Van den

Pol, 1980).

To keep these approximately 24 hour rhythms synchronized with the environment,

circadian rhythms need to be entrained by Zeitgebers. The most important and

potent Zeitgeber for the central SCN clock in mammals is light (Bell-Pedersen et al.,

2005). The light-input is perceived by specialized retinal ganglion cells in the eye

using a photo pigment called melanopsin (Berson et al., 2002). This photo pigment

differs from the rhodopsin in the rods and cones, used mainly for vision. The photic

signal is transmitted by the retinohypothalamic tract to the hypothalamus, which

projects to the SCN. Depending on the timing of the light exposure, the circadian

rhythm will delay or advance, resulting in e.g. waking up later or earlier, respectively

(Warman et al., 2003, Duffy et al., 1996). Whereas light can affect timing of the

clock in the SCN, timed food intake has been shown to entrain circadian rhythms

of peripheral clocks, independent of the SCN (Carneiro and Araujo, 2012). Social

cues are also sufficient to entrain human circadian rhythms (Aschoff et al., 1971).

Furthermore, melatonin is best known as output of the biological clock, but in turn

can also entrain or phase shift the clock in several species (Redman, 1997).

These circadian rhythms are endogenous, meaning that the rhythm persist under

constant conditions with a period of approximately 24 hours. The circadian rhythm

is self-sustaining and regulated on a molecular level by a molecular oscillator

composed of clock genes (described in Box 1). The circadian rhythm in the absence

of Zeitgebers such as light is called free running rhythm. In humans, the average



11

1free running period has been shown to be slightly longer than 24 hours (Czeisler

et al., 1999). In contrast, most mouse strains have an internal rhythm that is slightly

shorter than 24 hours, with variation between the strains (Pfeffer et al., 2015,

Schwartz and Zimmerman, 1990). As described previously, this internal rhythm is

synchronized with the environment by Zeitgebers. The relationship between the

timing of the biological clock and the timing of an external time cue is called the

phase angle of entrainment and determines chronotype; whether an individual is

an early type or a late type (Emens et al., 2009).

Box 1. Molecular clockAt the molecular level, the circadian rhythms are generated by an auto-regulatory

transcriptional-translation feedback loop (TTFL) (Lee et al., 2001). In short, the positive

limb of the TTFL consists of CLOCK and BMAL1 proteins, driving the transcription

of Period (Per1, Per2) and Cryptochrome (Cry1, Cry2) genes. In the negative limb,

PER and CRY proteins form a heterodimeric complex that translocates back into

the nucleus, repressing the CLOCK/BMAL1 driven transcription, thereby inhibiting

their own gene expression (Ko and Takahashi, 2006, Lee et al., 2001, Reppert and

Weaver, 2002). This molecular oscillator is coupled to output processes via clock-

controlled genes, including transcription factors (Reppert and Weaver, 2002).

Micro-array studies have revealed that approximately 10 % of the transcriptome is

under circadian control (Hughes et al., 2010, Miller et al., 2007, Panda et al., 2002).

Shift work-associated breast cancer riskWorking around the clock strains and disturbs this tightly regulated biological

clock, which eventually might result in adverse health events. Our 24/7 economy

demands people to work at irregular times and as a consequence, more and

more people are involved in atypical working schedules. Recent surveys in

Europe indicate that approximately 19% of the workers in the European Union

(EU) work at night and 17% are involved in shift work with permanent or rotating

shifts (Eurofound, 2012). In the Netherlands, approximately 1.2 million people

work sometimes or regularly during the night (CBS, 2010). Recent epidemiology

studies show that long-term shift work increases acute and chronic health risks,

like fatigue (Akerstedt and Wright, 2009), gastrointestinal complaints (Knutsson

and Boggild, 2010), and developing diabetes (Suwazono et al., 2006, Pan et al.,

2011) and cardiovascular disease (Gu et al., 2015, Fujino et al., 2006). Importantly,

work in shifts for many years was associated with an increased breast cancer risk


Chapter 1

12

in women (Akerstedt et al., 2015, Schernhammer et al., 2006). Epidemiological

studies concerning the relation between shift work or night work and breast

cancer risk have been summarized in several meta-analyses with contradicting

results (He et al., 2014, Ijaz et al., 2013, Jia et al., 2013, Kamdar et al., 2013, Megdal et

al., 2005, Wang et al., 2013). In the first meta-analysis Megdal et al. found a pooled

relative risk (RR) of 1.51 (95% CI, 1.36–1.68) for the association between shift work

and breast cancer, based on six studies (Megdal et al., 2005). Subsequent meta-

analyses reported lower RRs of approximately 1.2 (He et al., 2014, Jia et al., 2013,

Kamdar et al., 2013, Wang et al., 2013) or limited to no evidence for a shift work

breast cancer relationship (Ijaz et al., 2013).

Shift work is a complex combination of exposures, involving multiple aspects that

have been suggested to underlie the relationship between shift work and cancer

(Fritschi et al., 2011). These aspects include internal desynchronization, suppression

of melatonin or vitamin D levels, sleep disruption and lifestyle disturbances, all

shown in Figure 1. All suggested hypotheses are extensively discussed by Fritschi

et al. (Fritschi et al., 2011). In summary, the alternating shifts during shift work may

cause rhythms in peripheral function to become out of phase with the central

clock or sleep wake cycle, internal desynchronization. Additionally, both timing and

the degree of light exposure in shift workers is changed affecting both melatonin

and vitamin D levels. Light exposure during the night can result in melatonin

suppression (Davis and Mirick, 2006), which is suggested to increase breast cancer

development in experimental studies by various mechanisms (e.g. anti-oxidant

function) (Blask et al., 2005). Individuals working in night shift are suggested to

have less sunlight exposure, resulting in lowered vitamin D levels (Kimlin and

Tenkate, 2007), whereas sun exposure has been suggested to be protective of

several types of cancer (van der Rhee et al., 2013). Vitamin D has been suggested to

be one of the mediating factors in the preventive effect of sunlight on cancer (van

der Rhee et al., 2013), for example by inhibition of cell proliferation through various

mechanisms (Garland and Garland, 2006). Sleep disruption can cause a decrease in

melatonin levels, decreased immune activity and metabolic disturbance, which all

have been linked separately with increased cancer risk (Bovbjerg, 2003, Bianchini

et al., 2002, Blask et al., 2005). Lastly, lifestyle factors such as smoking, unhealthy

eating, age at first birth and duration of breast-feeding are possibly different for

shift-workers compared to day-workers and are all factors which increase cancer

risk.



13

1

Figure 1. Suggested scenarios underlying the relationship between chronic CRD and breast cancer (adapted from Fritschi et al. (Fritschi et al., 2011)).

Due to the observational nature of epidemiological studies, it is unknown which

aspect or combination of aspects is responsible for the observed increased health

risk in shift workers. Moreover, several factors can be either defined as part of the

exposure, intermediate factor or be a result of shift work exposure. For example,

significant differences in socio-economic status, smoking behavior, nulliparity,

hormone replacement therapy use, and obesity between women that had and

had not worked at night have been reported (Wang et al., 2012). An important

drawback of the epidemiological studies is that these studies require several years

to study effects on chronic diseases prospectively, and the ability to control the

aspects related to shift work is limited.

Additionally, exposure assessment of shift work is often lacking or incomplete due

to the retrospective nature of many studies and the diversity of shift work schedules

used. The first line of epidemiological studies, such as the Nurses’ Health Study

(Schernhammer et al., 2006), have been able to identify associations between shift

work and breast cancer using relatively crude exposure assessment metrics such

as ‘ever/never conducted shift work’ or ‘duration of working on a rotating schedule’.

Molecular epidemiology studies, as performed by Papantoniou et al. assessed the

exposure to shift work in more detail, reporting diurnal preference, light exposure,

circadian variation in melatonin, and sex hormone production (Papantoniou et al.,

2014). The findings of this cross-sectional study among day and night workers show

that night shift workers have decreased melatonin levels and differences herein

result from diurnal preference and intensity of light-at-night exposure. However,


Chapter 1

14

it will take many years before specific exposures during shift work are proven

to be causal for cancer risk in human studies. To study the causal relationship between circadian disturbance resulting from shift work and increased breast cancer risk, confounding factors should be excluded or taken into account and circadian disturbance exposure should be clearly defined or controlled.

Animal studies on the shift work-cancer relationshipAnimal studies allow detailed analysis of individual aspects of shift work, by

allowing the exclusion (melatonin, lifestyle factors) or measurement (internal

desynchronization, sleep disruption, vitamin D) of these individual factors.

Additionally, such studies can identify underlying changes in biological processes

that could provide targets for the development preventive measures. Animal

studies have a number of advantages compared to epidemiological studies. Firstly,

inter individual variation can be minimized in animal studies. Exposure models can

be developed to mimic human shift work allowing controlled uniform exposure,

in a genetically homogenous study population. Animal studies are therefore not

hampered by the environmental or lifestyle confounding factors, or exposure

assessment difficulties and therefore are better suited to study causality. Secondly,

in comparison to prospective cohort studies, the effects of shift work on long-

term health effect can be determined in a relatively short (1 year) time window.

Moreover, in shift work many environmental factors have changed, such as altered

light exposure, changed sleep-wake cycle (Grundy et al., 2009) and differently

timed food intake (Lennernas et al., 1995). In animal studies, these aspects can be

studied separately or in combination, providing insight into which disturbance(s)

of the circadian system is (are) (mostly) responsible for the observed health effects.

When using animals to study the relation between circadian rhythm disturbance

(CRD) and breast cancer risk, one can manipulate both sides of the equation:

circadian rhythm and breast cancer development. Disruption or disturbance of

circadian rhythms can be accomplished by alternating or shifting light schedules,

constant (dim) light exposure or disruption of the endogenous circadian rhythm.

The endogenous circadian rhythms can be disrupted through genetic modification

of the molecular clock, by mutating one of the core clock genes. Removing the

pineal gland will also result in the loss of circadian rhythmicity. In literature circadian

(rhythm) disruption is often used to describe manipulation of the circadian system.

In this thesis we differentiate between disruption and disturbance. The models



15

1in which the endogenous circadian system is modified either genetically or

anatomically, resulting in a completely or partially non-functional circadian system

are referred to as circadian disruption models. Whereas, circadian disturbance

refers to strain or stress imposed on the functional circadian system by exogenous

exposure, including but not limited to jet lag, altered timing of food intake and

day-time sleep.

Most studies make use of circadian disruption models or constant light exposure,

of which the majority shows an increase in breast tumor growth. The increased

tumorigenesis after disruption of the endogenous circadian rhythm or by the

“unnatural” constant light exposure model, which is known to desynchronize

mammalian clock neurons, resulting in behavioral and physiological arrhythmicity

(Ohta et al., 2005), emphasizes the relation between the circadian clock and tumor

development. However, altered light schedules provide a more realistic tool to

mimic constant disturbance of the circadian system resulting from shift work. Such

an approach has previously been employed by exposure to a nightly light pulse

(Travlos et al., 2001) or infusion with blood of women exposed to a nightly light

pulse (Blask et al., 2005). Whereas the first study found no effect on chemically

induced tumor growth, the latter showed a significant increase in MCF-7 xenograft

growth. Additionally, other circadian rhythm disturbance (CRD) models have been

described that mimic human shift work, including altered timing of 1) food intake,

2) activity and 3) sleep. A large variety of these models, as well as altered light

exposure, have been used to study the metabolic effects of human shift work with

a substantial number of indecisive results (Opperhuizen et al., 2015). Previously,

altered light schedules have shown to increase chemically induced liver tumor

growth (Filipski et al., 2009), but have not yet been employed in combination with

breast cancer induction models.

Since breast tumors are relatively rare in animals, studying breast cancer risk in wild

type animals requires a large number of animals. Therefore, breast cancer animal

models are used to reduce the number of animals needed. Breast cancer models

can be divided into chemically induced tumor development and xenografted

mammary gland tumors or tumor cell lines. Importantly, carcinogenesis comprises

three different stages, namely initiation, promotion and progression (Box 2 and

Fig. 2). The majority of the previous studies used chemically induced or xenografted

tumors provide evidence that CRD causes enhanced tumor growth and thus


Chapter 1

16

affect promotion and/or progression, but the studies could not give insight into

the effect of CRD on tumor initiation.

Box 2. CarcinogenesisThe multistage carcinogenesis theory is generally accepted and describes

three different stages between the initial carcinogenic stimulus and the final

manifestation of cancer (Berenblum, 1975). Following this theory, carcinogenesis

can be divided into initiation, promotion and progression. During the initiation

stage, a carcinogenic stimulus causes irreversible damage to the DNA, resulting in

potential for neoplastic development. The mutations causes either the activation

of a proto-oncogene or the inactivation of a tumor-suppressor gene. However,

without stimulation to proliferate the initiated cell will not develop into a

malignant tumor. The promotion stage is characterized by prolonged exposure to

promoting stimuli (Upton et al., 1986), causing the selective clonal expansion of the

initiated cell, resulting in a pre-neoplastic lesion. These cells become increasingly

unresponsive to cellular signals that regulate cell growth and proliferation. Further

genetic changes will result in more heterogeneity and aggressive characteristics

during the progression stage. The proliferation rate increases and tumor growth

becomes invasive, eventually resulting in distant metastasis. In this stage,

angiogenesis is required for tumor growth beyond 2 mm size (Folkman, 1985).

The circadian clock is involved in several aspects of carcinogenesis. DNA repair

pathways are in place to prevent irreversible damage to the DNA, resulting in

tumor initiation. It has been suggested that all aspects of the cellular response

to DNA damage are controlled or influenced by the circadian clock (Sancar et al.,

2010). Furthermore, it has been shown that the rate of tumor growth is highly

rhythmic during the day (You et al., 2005).

Overall, many animal experiments have been performed to elucidate the causal

relationship between circadian rhythm disturbance and mammary gland tumor

development (for references see Table 2). Table 1 summarizes all breast cancer

specific animal experiments published up to June 2015. The majority of the studies

used an exposure model resulting in disruption of the endogenous circadian

rhythm (n=13) or complete suppression of the circadian rhythm by continuous

light (n=11). Breast cancer induction in the majority of the studies involved

mouse models with rapidly growing tumors, either induced chemically (n=16) or

xenografted (n=5).



17

1

Figure 2. Carcinogenesis (Adapted from Weston & Harris Cancer Medicine 6th edition). Carcinogenesis comprises three stages: initiation, promotion and progression. In the initiation stage, a mutation occurs in a gene of one of the key regulatory pathways of the cell. Subsequently, the initiated cell selectively proliferates during promotion. Due to the genetic instability, additional mutations occur resulting in malignant cell growth: progression and ultimately to metastasis.

Overall, many animal experiments have been performed to elucidate the causal

relationship between circadian rhythm disturbance and mammary gland tumor

development (for references see Table 1). Table 1 summarizes all breast cancer

specific animal experiments published up to June 2015. The majority of the studies

used an exposure model resulting in disruption of the endogenous circadian

rhythm (n=13) or complete suppression of the circadian rhythm by continuous

light (n=11). Breast cancer induction in the majority of the studies involved

mouse models with rapidly growing tumors, either induced chemically (n=16) or

xenografted (n=5).

Table 1. Studies investigating the relationship between CRD and breast cancer development, shown as studies with positive outcome/all studies.

Circadian disruption / disturbance model

Breast cancer Induction

Altered light schedules

Continuous light exposure

Endogenous circadian rhythm disruption

Total

Spontaneous cancer development - 1/3 2/2 3/5

Chemically induced cancer 0/1 4/6 4/9 8/16

Xenografted tumors 1/1 2/2 2/2 5/5

Total 1/2 7/11 8/13 16/26

Although various animal models show increased tumor development or growth

upon circadian disruption, none of these studies fully recapitulate the human shift

work and carcinogenesis situation. For the circadian disruption models holds that


Chapter 1

18

the lack of a functional clock is different from the shift work situation where a

functional circadian clock is continuously strained or disturbed. Obviously, models

using xenografted tumors or tumor cell lines can only study tumor growth and

chemical induction of tumors is extremely potent, leaving limited space for tumor

initiation by CRD. Moreover, the window of exposure in these models is limited

to weeks due to the rapid tumor development, which does not allow mimicking

long-term exposure scenarios as experienced by shift workers. Interestingly, no

studies were performed combining spontaneous breast tumor development with

an exposure protocol focused on disturbing rather than disrupting the circadian

rhythm. Consequently, there is a need for studies that combine relevant models for CRD exposure and breast tumor development to provide experimental evidence for the shift work-cancer connection.

Li-Fraumeni mouse modelA mouse model with a predisposition for human relevant breast cancer provides

a unique tool to experimentally study the shift work breast cancer relationship.

Mutations in the tumor suppressor gene p53 are associated with increased tumor

development (Coles et al., 1992), progression (Borresen-Dale, 2003), recurrence

(Norberg et al., 2001) and decreased response to therapy. Additionally, Li-

Fraumeni syndrome patients, carrying germ line p53 mutations, are predisposed

to developing breast cancer at a relatively early age (Varley et al., 1997). Given

the apparent important role of p53 in preventing breast tumor development in

humans, a mouse model with a similar defect in p53 is a valuable tool to study

breast tumor development. Therefore, a transgenic mouse model was developed

with mammary gland specific expression of the p53.R270H mutation, resulting in

spontaneous mammary gland tumorigenesis mimicking human breast cancer

development (Wijnhoven et al., 2005). Expression of the mutation in mammary

tissue was achieved by crossing p53.R270H mutant mice with mammary-specific

Cre transgenic mice having Cre recombinase under the control of the hormone-

inducible Whey Acidic Protein (WAPCre mice; (Wagner et al., 1997)). This mouse

model has been shown to develop human relevant mammary gland tumors,

including hormonal receptor status, with a latency time of approximately one

year. This combination provides a human relevant endpoint and sufficient time for exposure to study the relationship between chronic CRD and breast cancer development experimentally.



19

1Ta

ble

2. D

etai

led

info

rmat

ion

of a

ll st

udie

s in

clud

ed in

Tab

le 1

.

Stud

y/au

thor

Stud

y po

pula

tion

Endp

oint

Brea

st c

ance

r in

duct

ion

Expo

sure

mod

elFi

ndin

gs

(Ani

simov

et a

l., 20

04)

Mic

eTu

mor

spe

ctru

mSp

onta

neou

sCo

nsta

nt (d

im) l

ight

Incr

ease

d tu

mor

inci

denc

e (K

haet

ski,

1965

)Ra

tsM

amm

ary

glan

d tu

mor

DM

BACo

nsta

nt (d

im) l

ight

Incr

ease

d tu

mor

inci

denc

e

(Ham

ilton

, 196

9)Ra

tsM

amm

ary

glan

d tu

mor

DM

BACo

nsta

nt (d

im) l

ight

Incr

ease

d tu

mor

inci

denc

e (K

otha

ri et

al.,

1984

, Mha

tre

et a

l., 19

84, S

hah

et a

l., 19

84)

Rats

Mam

mar

y gl

and

tum

orD

MBA

1)

Cons

tant

(dim

) lig

ht

2)

Pine

alec

tom

y1)

In

crea

sed

tum

or in

cide

nce

2)

No

effec

t(C

os e

t al.,

2006

)Ra

tsM

amm

ary

glan

d tu

mor

gr

owth

DM

BACo

nsta

nt (d

im) l

ight

Incr

ease

d tu

mor

gro

wth

(And

erso

n et

al.,

2000

)Ra

tsM

amm

ary

glan

d tu

mor

DM

BACo

nsta

nt (d

im) l

ight

Dec

reas

ed tu

mor

inci

denc

e(B

lask

et a

l., 20

03)

Rats

MC

F7 tu

mor

gra

ftTu

mor

cel

l lin

e gr

aft

Cons

tant

(dim

) lig

htIn

crea

sed

tum

or g

row

th(B

lask

et a

l., 20

05)

Rats

MC

F7 tu

mor

gra

ftTu

mor

cel

l lin

e gr

aft

Infu

sion

with

blo

od o

f wom

en

expo

sed

to a

nig

htly

ligh

t pul

seIn

crea

sed

tum

or g

row

th

(Sub

ram

ania

n an

d Ko

thar

i, 19

91)

Rats

Mam

mar

y gl

and

tum

orD

MBA

Pine

alec

tom

yIn

crea

sed

tum

or in

cide

nce

and

grow

th

(Bla

sk e

t al.,

1991

)Ra

tsM

amm

ary

glan

d tu

mor

NM

UPi

neal

ecto

my

No

effec

t(S

hah

et a

l., 19

84)

Rats

Mam

mar

y gl

and

tum

orD

MBA

Pine

alec

tom

yN

o eff

ect

(Lap

in, 1

978)

Rats

M

amm

ary

glan

d tu

mor

DM

BAPi

neal

ecto

my

Incr

ease

d tu

mor

inci

denc

e(A

uber

t et a

l., 19

80)

Rats

Mam

mar

y gl

and

tum

orD

MBA

Pine

alec

tom

yN

o eff

ect

(Tam

arki

n et

al.,

1981

)Ra

tsM

amm

ary

glan

d tu

mor

DM

BAPi

neal

ecto

my

Incr

ease

d tu

mor

inci

denc

e(T

ravl

os e

t al.,

2001

)Ra

tsM

amm

ary

glan

d tu

mor

NM

U1)

Pi

neal

ecto

my

2)

Nig

htly

ligh

t pul

se1)

N

o eff

ect

2)

No

effec

t(C

limen

t et a

l., 20

10)

Mic

eM

amm

ary

glan

d tu

mor

1)

DM

BA2)

M

MTV

-Neu

3)

Spon

tane

ous

Per3

kno

ck-o

ut1)

, 2),

3) In

crea

sed

tum

or in

cide

nce

(Yan

g et

al.,

2009

b)M

ice

MTC

L tu

mor

gra

ftTu

mor

cel

l lin

e gr

aft

Per2

kno

ck-d

own

Incr

ease

d tu

mor

gro

wth

(Yan

g et

al.,

2009

a)M

ice

MTC

L tu

mor

gra

ftTu

mor

cel

l lin

e gr

aft

Per1

kno

ck-d

own

Incr

ease

d tu

mor

gro

wth

(Vin

ogra

dova

et a

l., 20

09)

Rats

Tum

or s

pect

rum

Spon

tane

ous

Cons

tant

(dim

) lig

ht

Incr

ease

d tu

mor

gro

wth

, no

effe

ct o

n m

amm

ary

glan

d tu

mor

(Vin

ogra

dova

et a

l., 20

10)

Rats

Tum

or s

pect

rum

Spon

tane

ous

Cons

tant

(dim

) lig

htIn

crea

sed

tum

or g

row

th,

no e

ffect

on

mam

mar

y gl

and

tum

or(W

u et

al.,

2011

)Ra

tsM

CF7

tum

or g

raft

Tum

or c

ell l

ine

graf

tCo

nsta

nt (d

im) l

ight

Incr

ease

d tu

mor

gro

wth

(Bla

sk e

t al.,

2014

)Ra

tsM

CF7

tum

or g

raft

Tum

or c

ell l

ine

graf

tCo

nsta

nt (d

im) l

ight

Incr

ease

d tu

mor

gro

wth


Chapter 1

20

Monitoring CRD for interventionsGiven the wide variety of observed health effects, biomarkers are needed that

detect chronic CRD (long) before adverse health effects occur. Biomarkers are a

valuable source of information, enabling early or non-invasive detection of disease

or disease risk factors. Currently available or gold standard biomarkers for circadian

rhythm studies are melatonin, corticosterone and body temperature. These

markers are often used in studies to assess the effect of shift work (Niu et al., 2015,

Papantoniou et al., 2014, Sack et al., 1992), but have several limitations for the use in

large populations. An overall drawback of these classical markers of the circadian

rhythm is the need for around the clock measurement. In relation to this circadian

rhythmicity, for melatonin and body temperature the peak of the rhythm depends

on chronotype (Lack et al., 2009), which in turn alters with age, complicating the

use of these markers in longitudinal studies. Additionally, melatonin peak serum

levels can be suppressed by light exposure during the night (Davis and Mirick,

2006) making routine assessment more difficult. Moreover, classical markers are

acutely altered after impingement on the circadian system and are not related

to the (chronic) adverse health outcomes. Therefore, universal biomarkers for circadian rhythm disturbance are needed to compare CRD models, to evaluate the effect of preventive measures and to measure the personal risk of a shift worker.

Aim and outline of this thesisThe aim of this thesis is to study the causal relationship between chronic circadian

disturbance and adverse health effects. Additionally, identifying underlying

mechanisms and potential biomarkers for CRD to aid the development and study

of preventive measures are objectives as well. Finally, a new model to study health

effects of shift work in animal studies is presented.

In chapter 2 we used a unique breast cancer-prone mouse model to study the

causal relation between chronic CRD and breast cancer development. Besides

breast cancer risk, the effects of CRD on body weight gain, sleep probability and

classical circadian markers were investigated. As described previously, several

scenarios have been suggested to underlie the relationship between shift work and

cancer, in this chapter we show the relevance or irrelevance of these mechanisms.



21

1The presence of a causal relationship between shift work and adverse effects,

warrants the detection of biomarkers for early detection of disturbance in order

to develop preventive measures. Transcriptomic studies provide a valuable tool

to obtain a complete overview of gene activity in tissues of interest. In chapter 3,

a proof of principle is given using a comparative transcriptomics approach to

identify potentially blood detectable biomarkers for breast cancer. A similar

approach was used to identify biomarkers for CRD, independent of time of day,

in chapter 4. Besides using biomarkers for CRD, in large cohort studies one could

opt for the use of biomarkers related to disease outcome. The measurement of

these biomarkers in a molecular epidemiology-like setting requires knowledge

of their daily blood level patterns. Chapter 5 provides a descriptive study of the

daily variation of hormonal and lipid biomarkers, often used in large scale studies.

Development of evidence-based preventive measures further require insight into

underlying biological mechanisms.

In chapter 6, hepatic gene expression was analyzed to identify genes and

processes affected by chronic CRD. Both the effects on genes that show circadian

rhythmicity under normal conditions and on overall gene expression are described.

Subsequently, the role of these genes in biological processes relevant for the

observed phenotype was investigated.


CHAPTER 2

Chronically alternating light cycles increase breast cancer risk in

mice

Kirsten C.G. Van Dycke, Wendy Rodenburg, Conny T.M. van Oostrom,

Linda W.M. van Kerkhof, Jeroen L.A. Pennings, Till Roenneberg, Harry van Steeg

and Gijsbertus T.J. van der Horst

Current Biology, 2015, 25(14): 1932-7


Chapter 2

24

2

Summary

Although epidemiological studies in shift workers and flight attendants have

associated chronic circadian rhythm disturbance (CRD) with increased breast

cancer risk, causal evidence for this association is lacking (Pukkala et al., 1995,

Schernhammer et al., 2006). Several scenarios have been proposed to contribute

to the shift work-cancer connection: (i) internal desynchronization, (ii) light at

night (resulting in melatonin suppression), (iii) sleep disruption, (iv) lifestyle

disturbances and (v) decreased vitamin D levels due to lack of sun light (Fritschi

et al., 2011). The confounders inherent in human field studies are less problematic

in animal studies, which are therefore a good approach to assess the causal

relation between circadian disturbance and cancer. However, the experimental

conditions of many of these animal studies were far from the reality of human

shift workers. For example, some involved xenografts (addressing tumor growth

rather than cancer initiation and/or progression) (Blask et al., 2003, Filipski et

al., 2004), chemically induced tumor models (Hamilton, 1969, Shah et al., 1984)

or continuous bright light exposure, which can lead to suppression of circadian

rhythmicity (Anisimov et al., 2004, Wu et al., 2011). Here, we have exposed the

breast cancer-prone p53R270H©/+ WAPCre conditional mutant mice (in a FVB genetic

background) to chronic CRD by exposing them to a weekly alternating light-dark

(LD) cycle throughout their life. Animals exposed to the weekly LD-inversions

showed a decrease in tumor suppression. In addition, these animals showed an

increase in body weight. Importantly, this study provides the first experimental

proof that CRD increases breast cancer development. Finally, our data suggest

internal desynchronization and sleep disturbance as mechanisms linking shift

work with cancer development and obesity.


CRD increases breast cancer risk

25

2

Experimental Procedures

Experimental set-upTo study the effect of chronic circadian disturbance (CRD) on the development of

breast tumors, breast cancer-prone female p53R270H©/+WAPCre conditional mutant

mice in an FVB genetic background were chronically exposed to a LD-inversion

protocol. The p53 R270H mutation in human Li-Fraumeni patients occurs in

their germline and patients consequently develop tumors in a variety of tissues,

including breast cancer. Since we are specifically interested in breast tumor

development in the current study, we used the conditional p53R270H©/+ model, in

which the mutation was specifically activated in mammary gland tissue through

WAP-driven cre-recombinase. The generation of these mice has been previously

described (Wijnhoven et al., 2005). Cre recombinase expressing mammary gland

cells have been shown without pregnancy (Derksen et al., 2011), therefore virgin

mice were used in this study. At 8 weeks of age, mice were randomly assigned to

remain under a normal 12:12 hour light-dark (LD) cycle or to undergo a weekly

alternating 12:12 hour light-dark cycle. After approximately 18 shifts, on day 7 of

the last shift, a cross-sectional sample of animals was sacrificed around the clock,

with 4 hour intervals (cross-sectional study, n=4 per time point). Blood and tissues

were collected for further analysis. The remainder of the animals (longitudinal

study, n=25 per group) stayed under LD or weekly LD-inversion (CRD) conditions

and were sacrificed after when tumors reached approximately 1 cm3 or when

animals were found moribund. Due to premature animal loss, 21 animals per

condition were available for further analysis in the cross-sectional experiment.

In the longitudinal study, 20 and 21 animals were available for body weight and

tumor-free survival analyses for the control and CRD groups, respectively. In an

additional group of mice (n=5 per group), a radio transmitter (Physio Tel, TA11 TA-

F10; Data Sciences, St. Paul, MN) was implanted in the peritoneal cavity to record

locomotor activity and core body temperature every ten minutes.

The animal handling in this study was performed in compliance with national

legislation, including the 1997 Dutch Act on Animal Experimentation, and the

experiments were approved by the institute’s Animal Experimentation Ethical

Committee.


Chapter 2

26

2

Breast tumor developmentMice were palpated once a week to check for tumor development. The time of first

mammary tumor detection was registered and used to determine latency times. To

confirm the presence of a tumor and determine tumor type, paraffin-embedded,

formalin-fixed tumor sections were stained with H&E for histopathological

evaluation.

Body weight and food intakeMice and food were weighed weekly to determine body weight gain and food

intake. Body weight gain was expressed as percentages of body weight at the start

of the experiment and was statistically analyzed at 28 weeks of exposure, the last

time point without tumor-bearing animals. Animals were group-housed with three

to four animals per cage; therefore, food intake was expressed as average weekly

food intake. Food intake was measured between 8 and 15 weeks of exposure in

the cross-sectional study.

Gene expression, corticosterone and vitamin D levelsCircadian expression levels of clock genes Bmal1, Per1, Per2, Dbp and cell cycle

control gene c-Myc were determined in livers of control and weekly alternated

LD cycle exposed cross-sectional animals, using quantitative reverse transcription

polymerase chain reaction (RT-PCR). All oligonucleotide primers were obtained

from Life Technologies (Bleiswijk, The Netherlands). Total RNA was extracted from

RNAlater (Invitrogen, Grand Island, NY, USA) protected liver tissues using the

miRNeasy Mini Kit (Qiagen Benelux, Venlo, The Netherlands). Clock gene expression

was given as relative expression compared to a reference pool containing all

samples. Corticosterone and vitamin D serum levels were determined using ELISA

assays (Yanaihara Institute Inc. Shizuaka, Japan and Immunodiagnostics Systems,

Frankfurt am Main, Germany, respectively).

Body temperature and activityBody temperature and behavioral activity were recorded for 2 weeks at baseline,

after 1 shift and after 18 shifts. Cosine curves were fitted using the R statistical

software environment (http://www.r-project.org) to determine the acrophase of

activity and body temperature rhythms (i.e., peak time). Acrophase was expressed

in External Time (ExT) (Daan et al., 2002), with ExT 0 indicating mid-dark. Sleep

patterns were determined using mice telemetry data as described previously for



27

2

human wrist actimetry data (Juda et al., 2013b). In short, telemetric activity data

were collected in 10-min bins, and sleep episodes were automatically detected by

the method described earlier. Based on a non-rhythmic trend (calculated by using

centered moving 24-h averages), sleep (=1) and wake (=0) t were dichotomized

according to being below or above a selected threshold (15% of the trend). Onsets

and offsets of sleep episodes were then assessed by a correlation method as earlier

described in detail (Juda et al., 2013a, Roenneberg et al., 2015).

StatisticsAll data are expressed as means ± standard error of the mean (SEM) and were

visualized and statistically analyzed using GraphPad Prism software version 6.04

for Windows (GraphPad Software, San Diego California USA, www.graphpad.com).

Differences in body weight gain at 28 weeks, tumor latency time and vitamin D

serum levels were statistically tested using Kolmogorov-Smirnov test. Differences

between LD-inversion and LD animals in food-intake, gene expression and

corticosterone serum levels were analyzed using a two-way analysis of variance

(ANOVA) followed by Sidak’s posttest. Linear regression was performed to analyze

changes in total sleep, sleep in light and sleep in dark over time. Subsequently,

differences between the slopes for the LD and CRD group were analyzed. CircWave

Batch v5.0 software (Roelof Hut, www.euclock.org) was used to analyze circadian

rhythmicity of gene expression and serum levels. The ChronoSapiens software was

used to analyze behavioral data (Roenneberg et al., 2015).

Results

Classical circadian rhythmsTo investigate the potential causal links between chronic CRD and enhanced cancer

risk in more detail, we placed breast cancer prone p53R270H©/+ WAPCre conditional

mutant mice (further referred to as p53R270H©/+ WAPCre mice (Wijnhoven et al.,

2005) in a 12 hour light, 12 hour dark cycle (LD 12:12). At the end of every week,

the light or dark phase was extended to 24 hours to invert the LD cycle. Mice were

sacrificed after tumor development (longitudinal study) or around the clock after

approximately 18 LD-inversions (cross-sectional study). To monitor the extent to

which this protocol affected the circadian steady state, we recorded locomotor

activity and core body temperature (CBT) in an additional group of animals. Under


Chapter 2

28

2

baseline LD schedules, all p53R270H/+ WAPCre animals showed regular daily activity and

CBT rhythms (Fig. 1a for CBT; activity data not shown). The temperature maximum

is reached approximately at External Time (ExT) 20. After the LD-inversions, the CBT

rhythm re-established a stable phase of entrainment after 3-4 days of transients,

which appeared to be more gradual after the first inversion (Fig. 1b) compared

to inversion-week 18 (Fig. 1c). On day 7 after the 18th LD-inversion, representing

steady state, CBT showed a circadian rhythm, but its peak was significantly delayed

by 2 hours compared with age-matched control animals (p = 0.010).

Long-term health effectsIn the longitudinal experiment, as shown in Figure 2a, mice exposed to weekly

LD-inversions showed a larger increase in relative body weight compared to the

animals kept in a stable LD cycle (RM-ANOVA, group: p = 0.0319; time: p < 0.0001;

interaction: p < 0.0001). Although already apparent at week 6, group differences

only became significant after week 24 (Sidak’s posttest p < 0.05). The difference

in body weight gain between the groups did not reach significance in the cross-

sectional experiment (RM-ANOVA, group: p = 0.1410; time: p < 0.0001, interaction,

p = 0.4049), probably due to a shorter experiment time and fewer LD-inversions

(Fig. S1a). Differences in the amount of food-intake cannot explain the general

weight-gain in the LD-inversion groups, since we even found a small, significant

decrease in food intake in these animals compared to the stable LD controls (Fig.

S1b).

The latency to mammary gland tumor development was reduced by 17% in the

CRD-exposed mice compared to the LD control mice (Fig. 2b, median latency time:

42.6 versus 50.3 weeks, respectively; Kolmogorov-Smirnov p = 0.0127). Chronic

LD-inversion affected neither the number of tumor-bearing mice nor tumor

type (mammary gland tumors or other tumors). In both groups, approximately

80% of the animals developed mammary tumors, including carcinomas and

carcinosarcomas (Table S1).



29

2

Figure 1. Peak temperature phases under LD and CRD conditions. (A-C) of p53R270H/+WAPCre animals maintained under stable LD 12:12 conditions (closed symbols) or weekly alternating light cycles (open symbols) (n=5 animals per group) (a) before start of the light inversions, (b) at the first LD-inversion and (c) after 18 LD-inversions. Per graph, subsequent days are plotted from top to bottom. Time of day on the x-axis is expressed as External Time (ExT), with ExT 0 corresponding with mid-dark. The upper axis indicates the ExT before the LD-inversion. Values represent the mean ± SEM. Diamonds indicate the average temperature peak times of animals maintained under normal LD conditions. Data are presented as double plots to help visualizing phase shifts (day 0+1, day 1+2, day 2+3 etc. on consecutive lines). Grey areas indicate darkness.


Chapter 2

30

2

Figure 2. Long-term health effects resulting from CRD exposure. (a) Relative body weight gain of p53R270H/+WAPCre animals exposed to a regular LD cycle (closed symbols, n=20) or weekly alternating LD cycles (open symbols, n=21) in the longitudinal study. Note the significantly stronger weight gain of animals exposed to chronically alternating light cycles compared with animals maintained under a regular LD cycle in the longitudinal study (RM-ANOVA, group: F(1, 39) = 4.950, p=0.0319; time: F(27, 1053) = 42.48, p<0.0001; interaction: F(27, 1053) = 3.738, p<0.0001). Values represent the mean ± SEM. (b) Percentage of mice with palpable tumor in normal LD cycles (n=20; closed symbols) or chronic CRD conditions (n=21; open symbols). Black color indicates mammary gland tumor, whereas red color indicates other tumor types. See Table S1 for pathology data.

Investigating proposed mechanisms of shift work-related carcinogenesisTo gain more insight into these increased health risks resulting from chronically

alternating light cycles, we focused on the proposed mechanisms linking shift

work to cancer (Fritschi et al., 2011). We analyzed clock (Per1, Per2, Bmal1) and

clock-controlled (Dbp, c-Myc) gene expression in liver and corticosterone serum

concentrations (Fig. 3) to identify alterations and desynchronization among organ-

specific clocks and/or between central and peripheral clocks. In line with behavior

and CBT, Per1, Per2 and Dbp hepatic gene expression re-entrained within 7 days in

the new LD regime (tested after 18 LD-inversions; CircWave all p-values smaller than



31

2

0.05; Two-way ANOVA all p-values larger than 0.05). Only Bmal1 showed significant

interaction between group and time, clearly resulting from increased expression in

the CRD group at ExT10. In contrast to the unimodal 24-hr corticosterone rhythm

in control animals, we found a significant bimodal (12-hr) rhythm in the CRD mice

with a major peak at ExT 14 and a minor peak at ExT 2. We also analyzed the cell

cycle control gene c-Myc, which was previously suggested to be clock-controlled

and to play a role in accelerated tumor growth after chronic jet-lag (Filipski et al.,

2005, Fu et al., 2002). Although a slight phase advance appears to be induced by

CRD exposure, no significant differences in expression kinetics were observed

between the two groups.

We did not find an effect of CRD on 25-hydroxy-vitamin D levels (data not shown).

This might be well explained by the fact that laboratory animals are not exposed

to sun light, and accordingly are not stimulated to synthesize vitamin D.

To study whether chronically alternating light cycles caused sleep disruption, we

used the activity recordings to determine the total amount of predicted sleep.

As we estimated sleep based on periods of inactivity, in our study, sleep refers to

sleep probability rather than actual sleep. As shown in Figure 4 (left panels), the

total amount of sleep showed a significant increase both in CRD mice (p < 0.0001;

slope = 0.01543) and controls (p < 0.05; slope 0.003814) with increasing time in the

experiment. However, the increase over time was significantly different (p < 0.001)

between the two groups. While the controls increased their total sleep over the

course of the experiment by 10%, CRD mice slept 50% more. Under LD conditions,

the slight increase in sleep results mostly from sleep consolidation during the light

phase (upper middle panel). In contrast, CRD-exposed animals increased their

sleep in both light and darkness (lower left panels).

As to be expected from the frequent need to re-entrain after LD-inversions (see

also Fig. 1), sleep in light was reduced to half of baseline levels after the inversion

(day 0; Fig. S2, lower panels) and consequently increased to 150% in the dark

phase. However, while sleep in dark returned to approximately baseline after 4

days in the new regime, sleep in light almost increased to 175% of baseline. This

imbalance shows that the increase of total sleep shown in Figure 4 is mainly due

to overcompensation during the normal sleep times of this nocturnal rodent. LD

controls showed normal variations in sleep timing across the week (for separate

analyses of the recorded weeks, see Fig. S3a and Fig. S3b).


Chapter 2

32

2

Figure 3. Circadian expression of clock genes and c-Myc and corticosterone serum levels. Closed symbols = LD, open symbols = chronic LD-inversions. Circadian expression of clock genes Bmal1, Per1, Per2, Dbp and cell-cycle control gene c-Myc in liver and serum corticosterone levels (n= 3 or 4 mice per time point). Chronic LD-inversions did not significantly affect circadian expression of clock genes. Only minor differences were found at an individual time point for Bmal1 (*Sidak’s posttest p<0.05). The expression of c-Myc appears to show a phase advance in CRD-exposed animals, however no statistical differences were found. Corticosterone levels, however, were affected by the alternating light cycles. In contrast to the circadian rhythmicity in LD animals, corticosterone exhibits a significant 12-hr rhythm after 18 LD-inversions



33

2

with a major peak at ExT 14 and a minor peak around ExT 2 (p<0.05). Values at “lights on” were double plotted to help visualize circadian patterns. Values represent the mean ± SEM. Time of day on the x-axis is expressed as External Time (ExT), with ExT 0 corresponding with mid-dark.

Figure 4. Amount of predicted asleep per day (total sleep), in the light and dark phase, relative to baseline. Data represent means of animals per day, whereas linear regression analysis was performed on individual data points. Over time, total sleep and sleep in light significantly increased compared to baseline in both the LD (closed symbols) and LD-inversion (open symbols) group (Linear regression analysis: LD, total: p = 0.0492, light: p = 0.0002; LD-inversions, total: p < 0.0001, light: p < 0.0001). The amount of sleep in dark only increased after exposure to chronically alternating LD cycles (Linear regression analysis: LD: p = 0.5474 and LD-inversions: p = 0.0005). This gain of sleep was significantly larger for animals exposed to weekly LD-inversions compared to LD controls (comparison of slopes: total: F(1, 583) = 15.197, p = 0.0001; light: F(1, 583), p = 0.0391; dark: F(1, 583) = 8.744, p = 0.0032).


Chapter 2

34

2

Discussion

Human field studies only offer limited insights into potential causalities between

shift work and cancer due to the complex network of influencing variables,

such as different shift work schedules, genetic heterogeneity, or individual shift

work history (including the healthy workers phenomenon) – to name only a few

(IARC, 2010). The p53R270H©/+ WAPCre mouse model for spontaneous breast cancer

development allowed us to study the effects of chronically alternating light cycles

on breast cancer initiation and/or progression; thus, this model differs from previous

xenograft and chemically induced tumor models that do not reliably recapitulate

the human situation (Blask et al., 2003, Cos et al., 2006). To our knowledge, this

is the first study that unequivocally shows a link between chronic LD-inversions

and breast cancer development. It thereby provides experimental evidence for

previous epidemiological data associating shift work and jet-lag with increased

breast cancer risk. Epidemiological studies have also indicated an association

between social jet-lag, shift work and body weight, although conflicting data exist

(Kubo et al., 2011, Nabe-Nielsen et al., 2011, Roenneberg et al., 2012, Parsons et

al., 2014). Various mouse studies have shown that circadian disturbance induced

by continuous light, forced activity or a disrupted LD cycle causes an increase in

body weight (Coomans et al., 2013, Oishi, 2009, Salgado-Delgado et al., 2008). Our

study, enforcing circadian strain by using changes in the environmental LD cycle

supports the relationship between CRD and weight-gain. The finding that CRD-

exposed animals sleep more than controls, logically correlates with less activity

and could thus contribute to a stronger weight gain. But to what extent are the

results presented here for a causal link between CRD and increased breast cancer

risk in a nocturnal rodent also relevant to human shift work?

Shift work involves many aspects that could be involved in the causal mechanisms

that lead to increased health risks: internal desynchronization, melatonin

suppression due to light at night, sleep disruption, lifestyle disturbances (such as

smoking, [lack of ] breastfeeding, unhealthy diet, altered timing of food, etc.), and

decreased vitamin D levels due to lack of sunlight exposure (Fritschi et al., 2011).

The current study provides further insight into the relevance of these separate

aspects (for an overview: see Fig. S4). Melatonin suppression, decreased vitamin D

levels nor lifestyle disturbances (absent in mice) cannot account for the enhanced

cancer risk in animals kept under LD-inversion conditions: (i) p53R270H©/+ WAPCre

mice (FVB background) are melatonin-deficient; (ii) neither CRD mice nor controls



35

2

were exposed to sunlight, and we found no difference in 25-hydroxy-vitamin D

between groups. This is consistent with a study in night shift workers which found

no difference in vitamin D levels between fixed daytime workers, rotating shift

workers without night shift and rotating shift workers with night shift (Itoh et al.,

2011). (iii) Lifestyle factors did not differ (amount of food intake) in the LD-inversion

mice and the controls or were absent (e.g. breast feeding, smoking). Given the

combination of decreased breast tumor latency time and increased body weight

after chronic LD-inversions, it is tempting to speculate on the role of timing of food

intake. Potentially, changed timing of food intake disrupts metabolic processes,

resulting in adverse health effects. Previous studies have shown that timed feeding

can indeed (partially) rescue tumor and obese phenotypes (Filipski et al., 2005,

Fonken et al., 2010).

Desynchrony between the sleep-wake cycle and endogenous circadian

rhythmicity was recently shown to disrupt the circadian regulation of the

human transcriptome (Archer et al., 2014a) and suggested to be an important

factor underlying shift work-mediated breast cancer risk (Fritschi et al., 2013). In

the present study, 7 days after the last LD-inversion, we found a disturbance of

the circadian rhythmicity of corticosterone levels. It should be noted that blood

samples were taken under ketamine/xylazine anesthesia, known to slightly increase

corticosterone levels (Arnold and Langhans, 2010). However, as mice in the control

and LD-inversion group were treated in similar manner, anesthesia is unlikely to

explain the disturbed corticosterone rhythm in LD-inversed mice. In contrast to

the disturbed corticosterone rhythm, on day 7 after the last inversion, we did

not observe alterations in the daily patterns of hepatic clock gene expression,

indicating that the liver clock reentrained within one week. However, disturbance

of circadian rhythms at earlier time points is inherent of the re-entrainment to the

new LD cycle. Minor differences in c-myc expression could be indicative of larger

differences at earlier time points. This would be in line with other studies in which

phase advances induce c-myc expression directly and three days after exposure

(Filipski et al., 2005, Iwamoto et al., 2014). Previous studies have shown effects on

both corticosterone rhythm and liver clock gene expression by other circadian

disturbance protocols (Filipski et al., 2004, Barclay et al., 2012). Corticosterone

has been shown to be involved in the entrainment and resynchronization of

locomotor activity and peripheral clocks (Kiessling et al., 2010, Sage et al., 2004,

Sujino et al., 2012). Liver clock gene expression is more sensitive to input of feeding


Chapter 2

36

2

for entrainment (Sujino et al., 2012), which might explain the discrepancy and thus

desynchrony between the disrupted corticosterone rhythm and reentrained clock

gene expression in liver.

Sleep duration, sleep timing and sleep quality are central elements of human shift

work studies, both as outcome variables (in the form of descriptives of shift work-

related strain (Akerstedt, 2003) or post-intervention measures, (Neil et al., 2014,

Vetter et al., 2015)) and as a proposed countermeasure (e.g., schedule-specific sleep

strategies (Petrov et al., 2014)). As long-term sleep analysis in mice is challenging,

sleep was estimated based on inactivity, and accordingly refers to sleep probability

rather than actual sleep. However, there is a good consensus in the literature that

extended bouts of inactivity highly correlate with actual electroencephalogram

(EEG) and/or polysomnogram (PSG) detected sleep. In the Drosophila literature,

5-minute bouts of inactivity are considered sleep (Shaw et al., 2000) and

experiments in mice have shown that inactivity periods of 40 sec highly correlate

with EEG-scored sleep (Pack et al., 2007). Similarly high correlations are reported

for video-recorded immobility and EEG-scored sleep (Fisher et al., 2012). Since the

sleep analysis conducted here is based on relative immobility within 10-min bins,

we likely underestimate rather than overestimate the duration and frequency of

sleep episodes. Despite the potential risk of mistaking immobility in a wake mouse

for sleep, we consider our sleep-assessments as a good correlate for actual sleep

duration.

In human shift workers, sleep timing constantly changes, which results in shorter

sleep (and reduced sleep quality). In contrast, we observed that mice exposed

to chronic LD-inversions – not accompanied by forced sleep deprivation due to

work – sleep more. The finding that CRD-mice appear to sleep more than control

animals seems to contradict findings in shift-workers, who usually sleep less than

day-workers (Juda et al., 2013a). However, unlike real shift workers, the CRD protocol

in our animal study only involved changes in the LD cycle without the additional

sleep restrictions enforced by work-shifts impinging on usual rest times. Assuming

that the correlation between inactivity and sleep in mice is not influenced by the

exposure to chronically alternating light cycles, we propose that the increase in

amount of sleep observed upon CRD represents compensation for the constant

perturbations of circadian timing and sleep. Future studies, addressing sleep and

wakefulness by EEG rather than sleep probability, should provide a definite answer



37

2

to what extend sleep time and quality are affected upon chronically alternating

light cycles.

The search for the underlying mechanisms that link shift work and health detriments

(including increased cancer prevalence) is difficult in human shift work studies due

to the network of potentially mediating factors. Controlled laboratory experiments,

though performed in a nocturnal mouse model, allow the identification, isolation

and quantification of individual contributors. The present study provides additional

evidence for the role of internal desynchronization and sleep disruption in the

etiology of CRD-associated health risks and pathologies (e.g., obesity and cancer;

see Fig. S4). Female p53R270H©/+ WAPCre conditional mutant mice are ideal for

investigating this etiology for breast cancer risk in depth. Furthermore, the highly

controlled experimental conditions allow us to identify molecular biomarkers for

CRD or other (shift work-related) exposures. Despite being based on a nocturnal

mouse model, our results strongly suggest that individuals with hereditary (breast)

cancer predispositions should not be exposed to frequently changing Zeitgebers

as they exist for example in shift work or trans-meridian aviation. Due to the

growing 24/7 economy, shift work will become increasingly part of our society and

will, therefore, increasingly affect public health outcomes. Our experimental setup

provides a unique tool for exploring underlying mechanism as well as devising

countermeasures.


CHAPTER 3

Biomarker discovery using a comparative omics approach in a

mouse model developing heterogeneous mammary

cancer subtypes

Jeroen L.A. Pennings*, Kirsten C.G. Van Dycke*, Conny T.M. van Oostrom,

Raoul V. Kuiper, Wendy Rodenburg, and Annemieke de Vries

*joint first authors

Proteomics, 2012, 13: p2149-57


Chapter 3

40

3

Abstract

Identification of biomarkers for early breast cancer detection in blood is a

challenging task, since breast cancer is a heterogeneous disease with a wide

range of tumor subtypes. This is envisioned to result in differences in serum

protein levels. The p53R270H/+WAPCre mouse model is unique in that these mice

spontaneously develop both ER− and ER+ tumors, in proportions comparable

to humans. Therefore, these mice provide a well-suited model system to identify

human relevant biomarkers for early breast cancer detection that are additionally

specific for different tumor subtypes. Mammary gland tumors were obtained

from p53R270H/+WAPCre mice and cellular origin, ER, and HER2 status were

characterized. We compared gene expression profiles for tumors with different

characteristics versus control tissue, and determined genes differentially expressed

across tumor subtypes. By using literature data (Gene Ontology, UniProt, and

Human Plasma Proteome), we further identified protein candidate biomarkers for

blood-based detection of breast cancer. Functional overrepresentation analysis

(using Gene Ontology, MSigDB, BioGPS, Cancer GeneSigDB and proteomics

literature data) showed enrichment for several processes relevant for human

breast cancer. Finally, Human Protein Atlas data were used to obtain a prioritized

list of 16 potential biomarkers that should facilitate further studies on blood-based

breast cancer detection in humans.


Biomarkers for breast cancer

41

3

Introduction

Breast cancer is the most frequently diagnosed cancer and leading cause of cancer

death among women worldwide (Ferlay et al., 2010). The most efficient way to

reduce cancer mortality and morbidity is detection at an early stage, allowing

effective therapeutic intervention. One way to improve cancer detection could

be by means of a non-invasive test based on blood biomarkers. Preferably, these

serum markers detect breast cancer across multiple tumor types regarding

histopathology or diagnostic marker status, in an early stage of the disease.

Currently, CA15-3, CA27-29 and CEA are clinically used breast cancer serum

biomarkers for postoperative surveillance in patients with no evidence of disease

and monitoring therapy in patients with advanced breast cancer rather than early

detection (Sturgeon et al., 2008, Ludwig and Weinstein, 2005). Additionally, the

use of HER2/neu/ERBB2 (in shed form) in serum for disease monitoring in patients

without elevation of other tumor markers and for monitoring of Trastuzumab

treatment is under evaluation (Sturgeon et al., 2008), and CA125 has also been

suggested as tumor marker for advanced breast cancer (Norum et al., 2001).

However, these clinical serum biomarkers are only used for disease monitoring and

prognosis, and there are currently no FDA approved biomarkers for early detection

of breast cancer. Therefore, much international effort is put into identification and

validation of novel serum biomarkers for early breast cancer diagnosis (EDRN,

Kretschmer et al., 2011, Böhm et al., 2011, Opstal-van Winden et al., 2011).

Breast cancer is a heterogeneous disease, with a large variety of tumor subtypes

(Sorlie et al., 2001). Due to this heterogeneity, the identification and use of a

single biomarker for breast cancer detection is challenging. Here, we focus on

the identification of a panel of biomarkers which allows for detection with high

specificity and sensitivity across various kinds of breast cancer tumor types to

maximize screening applicability. In addition to a biomarker panel for all tumor

types, it would be beneficial to be able to distinguish between tumor subtypes,

as the prognosis and therapeutic approach is different for each subtype. Such

additional information on prognosis and therapy could lead to better follow-up and

referral of newly diagnosed patients. Relevant information in this respect would be

e.g. estrogen receptor alpha (ER) and HER2/neu status as well as histopathological

classification.


Chapter 3

42

3

A growing body of evidence demonstrate that mouse models for cancer

(summarized in (Kelly-Spratt et al., 2008)) show common molecular, biological

and clinical features compared to human cancers and as such these models are

considered a promising tool for identification of cancer biomarkers (Kelly-Spratt et

al., 2008). The human relevance of this approach is shown in a mouse model for

breast cancer where the overlap of the plasma proteome is compared to proteomics

studies of human breast cancer cell lines (Pitteri et al., 2008). However, mouse

models used thus far for breast cancer biomarker identification (Whiteaker et al.,

2007, Pitteri et al., 2008, Schoenherr et al., 2011, Whiteaker et al., 2011) exclusively

develop ER negative mammary gland tumors, while estrogen is thought to play

an important role in breast tumors development (Nandi et al., 1995). Moreover,

the majority of human breast tumors are ER positive. In our study, we use an

innovative mouse model developed previously in our laboratory (Wijnhoven et al.,

2005) which spontaneously develops ER positive and negative tumors, resembling

the human situation. This p53R270H/+WAPCre mouse model develops tumors of two

different histopathological types: carcinomas and carcinasarcomas. Carcinoma is

the most common tumor type found in women and therefore most relevant in

this study. Although carcinosarcomas are rare in humans, this subclass shows a

poor prognosis and early detection is highly necessary, since 5-year overall survival

reaches zero if detected at later stages (Gutman et al., 1995). In addition to the

estrogen receptor status, and the two histological subtypes, we were also able

to determine different HER2/neu status based on gene expression data. HER2/

neu amplification in human breast cancers is associated with lower 5-year survival

(Sotiriou and Pusztai, 2009). However, a positive HER2/neu status makes women

eligible for trastuzumab treatment which improves survival significantly (Gianni et

al., 2011). As such, HER2/neu based differentiation is highly important.

In this study, we use a genomics approach on this mouse model mimicking

heterogeneous breast tumor development in humans, to identify candidate

protein biomarkers relevant for early blood-based detection of human breast

cancer. First, we compare different types of tumor and control tissue from the

p53R270H/+WAPCre mouse model to determine genes which are informative across

multiple tumor subtypes versus control, or which are able to distinguish between

tumor subtypes. In the next step, we determine which of these mouse genes have

a human equivalent that encodes for a protein potentially detectable in human

serum. Finally, we compare the obtained lists with relevant information in other



43

3

databases to functionally characterize and prioritize the proteins for follow-up

studies.

Materials and Methods

Animal experimentThis study was approved by the Animal Experimentation Ethical Committee of

our institute. Animal handling in this study was carried out in accordance with

relevant Dutch national legislation, including the 1997 Dutch Act on Animal

Experimentation.

Generation of p53R270H/+WAPCre mice and the phenotypic characteristics have

been described previously (Wijnhoven et al., 2005). Heterozygous p53R270H

and WAPCre mice, originally in a hybrid background (Wijnhoven et al., 2005),

were crossed to FVB mice to generation F6. Female p53R270H/+WAPCre mice

were generated by crossing female heterozygous p53R270H mice with male

heterozygous WAPCre mice. All mice were weighed weekly and checked for the

development of tumors by palpation. Mice were sacrificed immediately after

detection of a palpable tumor (62±12 weeks of age), since our analysis was aimed

at the identification of biomarkers for early detection of breast cancer. All tumors

were dissected and weighed, the tumor size ranged from 52 to 988 mg with a

median tumor weight of 328 mg. Control mammary glands were isolated from

the same animals at the opposite site of the tumor. Tumors were processed for

histopathology following standard procedures. For comparison with control tissue

from a wild type strain, female FVB/N mice were sacrificed at 55 weeks of age and

mammary tissue was collected.

Histology and immunohistochemistryParaffin-embedded, formalin fixed tumor sections were stained with H&E for

histopathological evaluation. ER expression was analyzed using the rabbit

polyclonal antibody MC-20 (1:200; Santa Cruz Biotechnology, Santa Cruz, CA). In

this study we found that 74 percent of the mammary tumors showed ER positivity,

which is comparable to earlier studies with this mouse model and the human

situation (Wijnhoven et al., 2005).


Chapter 3

44

3

Erbb2/HER2/neu statusIn clinical diagnosis, mRNA expression levels have been shown to be a reliable

method for the determination of HER2/neu status of breast tumors (Capizzi et al.,

2008). Therefore, we used our microarray expression data for Erbb2 (the murine

form of HER2) to determine the HER2/neu status of the mammary gland tumors

in our study. We used a cutoff value of mean log transformed expression of normal

mammary gland tissue plus 3 standard deviations (Capizzi et al., 2008) to define

which tumors were HER2/neu positive. In our data, this corresponded to a three-fold

up-regulation of Erbb2 compared to control tissue. Using this cutoff, we found that

21 percent of the mammary tumors were HER2/neu positive, which is close to the

average reported value of 26% in human breast cancer (Revillion et al., 1998). Please

note that we will use the phrase HER2 positive/negative throughput the paper to

relate murine Erbb2 gene expression to the terms used in human clinical diagnostics.

Microarray analysisRNA was extracted from control mammary glands and mammary tumors using the

miRNeasy Mini Kit (Qiagen Benelux, Venlo, The Netherlands). RNA concentrations

were measured using a NanoDrop spectrophotometer (NanoDrop Technologies,

Wilmington, DE, USA). The integrity of the RNA samples was determined with the

BioAnalyzer (Agilent Technologies, Amstelveen, The Netherlands) using the RNA

nano 6000 kit (Agilent Technologies). RNA was further processed for hybridization

to Affymetrix HT Mouse Genome 430 PM Array Plates at the Microarray Department

of the University of Amsterdam, the Netherlands. RNA amplification, labeling and

genechip hybridization, washing and scanning were carried out according to

Affymetrix protocols.

Altogether, 30 samples were used for further analysis: 6 control FVB, 5 control

p53R270H/+WAPCre (Control TG), 9 carcinomas (Car) ER+, 2 carcinomas ER−, 5

carcinosarcomas (Csc) ER+, and 3 carcinosarcomas ER−. Of the 9 ER+ carcinoma

samples, 4 were HER2/Neu positive; all other samples were HER2/neu negative.

Data analysisAffymetrix CEL files were normalized using the Robust Multichip Average

(RMA) algorithm and the MBNI custom CDF (http://brainarray.mbni.med.umich.

edu/ Brainarray/ Database/ CustomCDF/) for this chip. RMA output consisted

of 16492 probe sets corresponding to unique Entrez GeneIDs. Quality control



45

3

consisted of visual inspection of residual plots and other RMA QC parameters.

Further data analysis such as ANOVA and Principal Component Analysis (PCA) was

carried out using R software unless indicated otherwise. Gene expression data

between (two or more) experimental groups were compared using ANOVA, where

p values < 0.001 and False Discovery Rate (FDR) < 5% were considered significant.

Additionally, for tumor vs. control comparisons, only genes up-regulated in the

tumor group(s) were considered in further analyses, as the aim of this study was

to detect tumor-originating markers for early cancer detection in serum-based

diagnostics. By making further Boolean combinations of regulated gene lists, we

obtained genes informative across or between tumor subtypes.

To determine which genes code for proteins potentially detectable in human

serum, we determined which proteins have a human equivalent that is annotated

in Gene Ontology as extracellular, in UniProt as secreted, or has been experimentally

detected in plasma or serum as part of the Human Plasma Proteome project

(Anderson et al., 2004).

Functional annotation and overrepresentation analysis was carried out using

DAVID (http://david.abcc.ncifcrf.gov) (Huang et al., 2009) for generally known

databases such as Gene Ontology and UniProt. Further gene set overrepresentation

analyses were performed using an in-house developed tool (based on the DAVID

methodology) for other custom functional gene set libraries based on BioGPS tissue

expression data (www.biogps.org, (Wu et al., 2009)), MSigDB-C2 gene sets curated

from genomics literature (www.broadinstitute.org/gsea/msigdb/, (Liberzon et al.,

2011)), MSigDB-C4 cancer gene expression modules (Segal et al., 2004) the cancer

signatures GeneSigDB (http://compbio.dfci.harvard.edu/genesigdb/, (Culhane et

al., 2010)), and an in-house made collection of protein lists from mouse blood-

based proteomics literature (Pitteri et al., 2008, Whiteaker et al., 2011, Whiteaker et

al., 2007, Schoenherr et al., 2011).

Protein expression in human breast cancer and normal mammary tissue was

compared using the Human Protein Atlas (www.proteinatlas.org) (Ponten et al.,

2008). This site contains semi-quantitative (absent – weak – moderate – strong)

data. We considered a protein up-regulated in human tissue if at least 50% of the

tumor samples in the database displayed an expression level at least one level

higher than that of normal mammary gland.


Chapter 3

46

3

A visual summary of the workflow used in the data analysis is given in Figure 1.

Figure 1. Summary of the data analysis workflow, with cross-references to corresponding table and figures, and a summary of the functional enrichment analysis results.



47

3

Results

Principal Component AnalysisPrior to statistical analysis, whole genome transcription profiles were visualized in a

PCA (Figure 2). This revealed that, on the whole, tumor samples are clearly distinct

from control samples. Additionally, the PCA indicated that several tumor subclasses

with regards to histology and receptor status could already be distinguished (e.g.

carcinoma HER+, carcinoma ER−, carcinosarcoma ER+ versus ER−), whereas the

two control groups appeared highly similar. This latter finding was confirmed by

statistical analysis: there are no genes with significant expression changes between

the two mammary tissue controls groups (Table 1) and therefore these groups

were combined in further analyses.

Figure 2. Principal Component Analysis on tumor and control tissue gene expression data.

Similarities in genes expression changes across tumor subtypesWe compared gene expression changes between controls and tumor samples

for either all tumors or by subset based on histopathology, ER, or HER2 status

(Table 1, Figure 3). Since this study was set up to specifically identify tumor-

originating markers for cancer detection by means of a blood-based test, we not


Chapter 3

48

3

only considered statistical significance (p < 0.001 and FDR 5%), but also set the

demand that genes are up-regulated in tumor vs. control. After calculating the

number of significantly up-regulated genes in tumor groups, we also determined

how many of these have a human equivalent that is potentially blood-detectable,

again with eventual screening implementation in mind. Numbers of genes

meeting these criteria are given in Table 1. For tumor (subtype) versus control

comparisons, we found between 2185 and 3471 up-regulated genes, of which

238 to 458 (on average 11%) were potentially detectable in human blood. The fold

ratio for these up-regulated genes ranged from 1.21 to 39.71 (median 1.85), for the

potentially blood-detectable markers the fold up-regulation ranged from 1.28 to

39.71 (median 2.21).

Table 1. Statistical comparisonsa and further combinations

Query Comparison Regulated genes Of which human blood detectable

A Control FVB <> Control TG 0 0

B all tumor > all control 2850 313

C carcinoma > all control 2330 238

D carcinosarcoma > all control 3471 458

E ER+ > all control 2768 309

F ER− > all control 2185 247

G HER2+ > all control 2768 307

H HER2− > all control 2749 313

J carcinoma <> carcinosarcoma 350 76

K ER+ <> ER− 0 0

L HER2+ <> HER2− 178 24

Combination Description Regulated genes Of which human blood detectable

C AND D carcinoma and carcinosarcoma vs control 1942 200

(C OR D) AND J carcinoma/carcinosarcoma differentiation 197 51b

C AND J specific for carcinoma 53 11

D AND J specific for carcinosarcoma 171 48

(C AND D) AND J applicable for both 27 8

E AND F ER+ and ER− vs control 1815 202

(E OR F) AND K ER+/ER− differentiation 0 0

G AND H HER2+ and HER2− vs control 1887 203

(G OR H) AND L HER2+/HER2− differentiation 117 15c

G AND L specific for HER2+ 109 14

H AND L specific for HER2− 25 2

(G AND H) AND L applicable for both 17 1

ALL (B,C,D,E,F,G,H) Shared across all subtypes 1395 149d

a based on p < 0.001 and FDR < 5%b Supplementary Table 2c Supplementary Table 3d Supplementary Table 1



49

3

Next, we determined the overlap between the various gene lists. We noticed that

for each of these lists, roughly half (37% - 74%) was present in an overlap with

other sets (Table 1, Figure 3). This indicated that there is a common set of genes

that discriminates the various tumor groups from control tissue, but not the sub-

tumor types by themselves. The most comprehensive overlap across all of the

comparisons resulted in 1395 regulated genes of which 149 met the criteria for

protein blood detectability. These 149 genes are listed in Supplementary Table

1. Fold up-regulations ranged from 1.30 to 16.80 (median 2.13) for the set of

1395 genes, and from 1.39 to 16.34 (median 2.36) for the 149 potentially blood-

detectable genes. As for the single tumor (subtype) versus control comparisons,

there is a trend in that the potentially blood-detectable subset shows larger fold

up-regulations.

Figure 3. Venn diagram for the various comparisons made. Overlap of discriminative genes based on histological classification (red), ER status (green), HER2 status (blue), and across all tumor subtypes (bold).

GO functional annotation analysis on this list revealed enrichment for terms related

to the cytoskeleton, extracellular matrix (including remodeling), cell division,

(negative regulation of ) apoptosis, and immune/inflammatory response (Figure 1).


Chapter 3

50

3

Enrichment analysis on the MSigDB-C2 database showed mainly significance (FDR

5%) for gene sets related to cancer, including several breast cancer sets (Poola

et al., 2005, Schuetz et al., 2006, Sotiriou et al., 2006), with other significant gene

sets being mainly involved in stress and immune responses. Enriched MSigDB-C4

cancer modules were related to cytoskeleton, extracellular matrix, cell division, and

inflammatory response. The BioGPS-derived tissue-associated gene set library did

not reveal a match between normal mammary tissue and the shared 149 marker

set, but instead showed enrichment for predominantly immune set-related

signatures, perhaps reflecting the endogenous response to the tumor (Figure 1).

Comparison to literature mouse plasma breast cancer proteomics sets revealed

one significant enrichment, namely to Table S6g of Whiteaker et al (Whiteaker et

al., 2011) which describes their verification of plasma biomarkers on individual

mice. This overlap comprised six genes (Hsp90b1, Hspa5, P4hb, Pfn1, Tnc, Ywhaz).

Finally, the cancer GeneSigDB was used to determine if there was significant

overlap to cancer literature gene sets. The analysis showed significant enrichment

for 171 of the 2142 cancer signatures. These signatures were derived from several

tissues, although the fraction of significant breast cancer signatures (63 enriched

out of 476 breast cancer sets) was relatively highest, being 1.7 times that for the

overall (171 / 2142) significant fraction. The most significant breast cancer gene

set was that by Lauss et al (Lauss et al., 2008) containing consensus genes from

literature studies to predict breast cancer recurrence. Next, we determined which

genes are most characteristic for the 63 enriched breast cancer signatures versus

the rest of the database. We found that this enrichment could mainly be ascribed

to genes involved in cell division. Overall, the high match to literature gene sets

suggests our marker set has considerable resemblance to marker sets found in

(predominantly) human breast cancer.

Differences between tumor subclassesBesides biomarkers common to all tumor subclasses, we also set out to find

biomarkers that might help to discriminate between various tumor subclasses.

Indeed, the PCA in Figure 2 indicated this would be possible. For the carcinoma

vs. carcinosarcoma comparison we found 350 differentially expressed genes, 76

of which are potentially human blood-detectable (Table 1, Figure 3). Since latter

implementation would only be practical if a biomarker was increased in either of

the two subclasses when compared to a control, we determined the overlap of

the carcinoma vs. carcinosarcoma genes with the sets of genes up-regulated in



51

3

carcinoma and/or carcinosarcoma to control. This led to 51 genes, which are listed

in Supplementary Table 2. As can be seen from this table, 48 of these genes show

higher expression in carcinosarcoma than in carcinoma. Additionally, GO functional

enrichment analysis showed enrichment for extracellular matrix, extracellular

matrix remodeling, and angiogenesis. These findings are in line with the more

aggressive clinical phenotype of this subtype. Of the 51 genes in Supplementary

Table 2, 8 overlap with the set of 149 shared genes (Supplementary Table 1),

namely Actn1, Inhba, Msr1, Ncam1, Plau, Pls3, Tpm4, and Wisp1. Enrichment analysis

on C2, C4 and BioGPS datasets found significance for (breast) cancer, extracellular

matrix (remodeling), angiogenesis, and immune response gene sets (Figure 1).

No significant overlap with literature mouse plasma cancer proteomics sets was

found. In the GeneSigDB sets there was enrichment for 79 sets, 26 of which were

breast cancer related. Here the breast cancer-specific enrichment was mainly due

to extracellular matrix (remodeling) associated genes.

When the analogous comparison was made for ER+ and ER− tumors vs. each

other and the control, no differentially expressed genes were found. Apparently,

following the genomics strategy described here, no potential biomarkers are

identified to discriminate between ER+ and ER− breast cancers in human blood.

For the HER2 status, 178 genes were differentially expressed between HER2+

and HER2− tumors, 117 of which were up-regulated in either subtype to control

(Table 1, Figure 3). Of these, 15 were potentially blood-detectable (Supplementary

Table 3). Of these, 14 were higher expressed in HER2+ tumors, including, as could

be expected, the Erbb2 gene itself. Mutual comparisons showed no overlap

between this list and the list of 149 markers shared across tumor subtypes versus

control, nor the 51 genes discriminating carcinoma from carcinosarcoma. No GO

enrichment was found for these 15 genes. Additional enrichment analyses only

showed enrichment for one gene set across all five databases, namely the C2 entry

for an amplification hot spot in the human chromosome 17q11.1-q21; 17q25

region described by Myllykangas et al (Myllykangas et al., 2006), which indeed

contains the (human homolog of the) Erbb2 gene (Figure 1).

Further prioritization of potential biomarkers

To translate our results obtained using a mouse model to the human situation

and prioritize the markers for further human follow-up studies, we used the data


Chapter 3

52

3

available on the Human Protein Atlas to compare protein expression in human

breast cancer tissue and normal mammary gland for the 149 shared genes in

Supplementary Table 1. For the 116 proteins where such human information was

available, 16 showed an increased expression in human cancer tissue compared

to normal mammary gland. These 16 proteins are given in Table 2. Their gene

expression fold up-regulations ranged from 1.81 to 7.28 (median 2.13) with

standard deviation values (see also Supplementary Table 1) being approximately a

third of the fold up-regulation (median relative SD 36%). A relatively high standard

deviation (5.62, 77% relative SD) was found for VCAN. Functionally, the majority of

these 16 proteins are associated with the cytoskeleton.



53

3

Tab

le 2

. Prio

ritiz

ed s

hare

d se

rum

mar

kers

, bas

ed o

n hu

man

tiss

ue e

xpre

ssio

n da

ta

Hum

an G

eneI

DH

uman

sym

bol

Des

crip

tion

Ave

rage

ratio

Tu

mor

/Con

trol

aH

uman

pro

tein

sta

inin

g: m

edia

n br

east

can

cer v

ersu

s no

rmal

bre

ast t

issu

eb

60A

CTB

actin

, bet

a2.

00m

oder

ate

vs w

eak

7987

9CC

DC

134

coile

d-co

il do

mai

n co

ntai

ning

134

1.81

mod

erat

e vs

wea

k

1072

CFL

1co

filin

1 (n

on-m

uscl

e)2.

45m

oder

ate

vs w

eak

2146

EZH

2en

hanc

er o

f zes

te h

omol

og 2

(Dro

soph

ila)

2.92

mod

erat

e vs

neg

ativ

e

2316

FLN

Afil

amin

A, a

lpha

2.10

wea

k vs

neg

ativ

e

1090

5M

AN

1A2

man

nosi

dase

, alp

ha, c

lass

1A

, mem

ber 2

2.33

stro

ng v

s m

oder

ate

4627

MYH

9m

yosi

n, h

eavy

cha

in 9

, non

-mus

cle

2.09

mod

erat

e vs

wea

k

4830

NM

E1no

n-m

etas

tatic

cel

ls 1

, pro

tein

(NM

23A

) exp

ress

ed in

2.27

stro

ng v

s m

oder

ate

1133

3PD

AP1

PDG

FA a

ssoc

iate

d pr

otei

n 1

2.72

mod

erat

e vs

neg

ativ

e

2910

8PY

CA

RDPY

D a

nd C

ARD

dom

ain

cont

aini

ng2.

17st

rong

vs

mod

erat

e

2606

4RA

I14

retin

oic

acid

indu

ced

143.

77w

eak

vs n

egat

ive

5491

0SE

MA

4Cse

map

horin

4C

2.40

wea

k vs

neg

ativ

e

7168

TPM

1tr

opom

yosi

n 1

(alp

ha)

2.59

wea

k vs

neg

ativ

e

5159

2TR

IM33

trip

artit

e m

otif

cont

aini

ng 3

32.

12m

oder

ate

vs w

eak

1462

VCA

Nve

rsic

an7.

28w

eak

vs n

egat

ive

7511

XPN

PEP1

X-pr

olyl

am

inop

eptid

ase

(am

inop

eptid

ase

P) 1

, sol

uble

2.01

wea

k vs

neg

ativ

e

a add

ition

al v

alue

s fo

r tum

or s

ubcl

ass

ratio

ave

rage

s an

d st

anda

rd d

evia

tions

can

be

foun

d in

Sup

plem

enta

ry T

able

1b re

trie

ved

from

Hum

an P

rote

in A

tlas

(ww

w.p

rote

inat

las.o

rg)


Chapter 3

54

3

Discussion

Breast cancer is the leading cause of cancer death worldwide (Ferlay et al., 2010).

Early diagnosis of cancer is necessary for improvement of survival and reduction

of morbidity. Therefore, there is a need for sensitive testing strategies specifically

for early detection. For improvement of current early detection rates, blood-

based diagnostics could be a valuable addition to (or next to) existing screening

programs, because of their low costs and flexible logistics. The first step towards

implementing such blood-based detection is the identification of biomarkers that

detect across multiple tumor types, to increase overall sensitivity. A secondary step

would be the identification of biomarkers that are able to differentiate between

tumor subtypes to improve follow-up and referral of newly diagnosed patients,

i.e. a more personalized approach. In the study described here, we use a genomics

approach on tumor tissue compared to normal mammary gland to identify

protein candidate biomarkers relevant for blood-based early detection of human

breast cancer. Interestingly, we hereby differentiated the mammary tumors with

regards to histological subtype and receptor status, to reflect human variation in

these tumor characteristics. Moreover, we used an experimental set up specifically

aimed at the early detection of cancer, rather than a set up with more progressed

cancers.

Mouse models are frequently applied in biomarker discovery research (Kelly-

Spratt et al., 2008). Earlier studies have shown that although both human and

mouse breast tumors have a heterogeneous profile, similarities can be found and

therefore, mouse models are useful in human breast cancer biomarker discovery

studies (Klein et al., 2007, Pitteri et al., 2008). In recent years, some studies have

focused on early detection, such as a plasma proteomics study between PyMT

transgenic tumor-bearing mice and matched controls by Pitteri et al. (Pitteri et al.,

2008), and several between tumor and normal mammary tissue from a conditional

HER2-driven mouse (Schoenherr et al., 2011, Whiteaker et al., 2011, Whiteaker et

al., 2007). However, these literature studies are limited in number, and none have

taken into account the differentiation between tumor subtypes within a single

mouse model. Here, we first report the use of a humanized mouse model for breast

cancer that represents the heterogeneity of breast cancer, within an identical

genetic background. Two different histological subtypes, and ER as well as HER2

positive and negative tumors are represented in this model. We did not determine



55

3

progesterone receptor (PR) status, as clinically PR expression has little prognostic

value and is merely used as an additional factor next to ER for the response to

hormonal therapy (Sotiriou and Pusztai, 2009).

In this study we identified 149 potentially blood-detectable biomarkers

(Supplementary Table 1). The identified genes were significantly up regulated in all

tumor types compared to normal mammary gland tissue and potentially blood-

detectable according to GeneOntology, Uniprot or the Human Plasma Proteome

Project. Moreover, we compared protein expression in human breast tumors and

normal breast tissue using the Human Protein Atlas. With these steps we translated

mouse mammary gland tumor gene expression to 16 prioritized candidate blood-

detectable human breast cancer markers (Table 2).

Our biomarker panel is enriched for several biological functions, such as

cytoskeleton, extracellular matrix (including remodeling), cell division, (negative

regulation of ) apoptosis, and immune/inflammatory response (Figure 1). These

terms are related to cancer processes in general, although not specifically to breast

cancer. This is corroborated by the enrichment for MSigDB-C2 and GeneSigDB

signatures for breast- but also other types of cancers. If the intended use of the

biomarker panel would be specifically breast cancer, biomarkers specific for breast

cancer (or other tumors) should also be included. Additionally, it can be noted that

several immune/inflammatory response markers can also be triggered by other

diseases and such markers would need to be combined with non-immunological

markers to provide sufficient specificity for (breast) cancer.

We did not detect any significant differential expression for the clinically used

markers CA15-3, CA125, CEA, and CA27-29. This can probably be attributed to

the fact that – in line with our aim of finding early detection biomarkers – we

dissected mice as soon as a palpable tumor was detected. These clinical markers

are more often found at more advanced tumor stages and are considered more

useful for tumor monitoring and prognosis rather than early detection (Ludwig

and Weinstein, 2005, Zeiwar et al., 2007, Molina et al., 2010, Tampellini et al., 2006).

We identified more potential biomarkers across tumor types than we did for

discrimination between tumor types. As mentioned before, breast cancer is a

heterogeneous disease and in this study several tumor types were represented. It


Chapter 3

56

3

seems that despite this heterogeneity there are still substantial similarities across

tumors types, which involve known cancer deregulated processes such as cell

division and tissue invasion. This would be advantageous to finding biomarkers to

improve current breast cancer screening. In the future, additional differentiation

between tumor subtypes based on tumor type specific biomarkers or genetic

make-up may be beneficial.

Although our mouse model represents the heterogeneity found among human

breast tumors and the regulated genes found in this study are involved in

mechanisms relevant for human breast cancer, the use of any mouse model for

human breast cancer has inherent limitations and application of our biomarker

panel in a large-scale screening setting would require several validations steps.

Therefore, naturally, our potential biomarker panel needs to be tested in human

blood samples, to determine the usefulness, and subsequently sensitivity and

specificity of this panel for early detection of breast cancer. The main potential

limitation in human samples is that the variation between individuals is much

larger and statistical power for early tumor detection would decrease considerably.

In fact, studies have shown that it is even difficult to find reproducible markers

within a clinically homogenous group of patients due to the individual variations

and heterogeneities (Ein-Dor et al., 2005). Further studies in small human cohorts

or humanized mouse models, using multi-marker assay methods such as Luminex

or antibody arrays are therefore needed to determine which markers in our panel

are sufficiently robust for eventual testing on a larger human cohort.


3


CHAPTER 4

Biomarkers for circadian rhythm disturbance independent of time

of dayAdapted from:

Biomarkers for circadian rhythm disruption independent of time of day

Kirsten C.G. Van Dycke, Jeroen L.A. Pennings, Conny T.M. van Oostrom,

Linda W.M. van Kerkhof, Harry van Steeg, Gijsbertus T.J. Van der Horst,

Wendy Rodenburg

PLoS One, 2015, 18;10(5)


Chapter 4

60

4

Abstract

Frequent shift work causes disturbance of the circadian rhythm and might on

the long-term result in increased health risk. Current biomarkers evaluating the

presence of circadian rhythm disturbance (CRD), including melatonin, cortisol

and body temperature, require 24-hr (“around the clock”) measurements, which is

tedious. Therefore, these markers are not eligible to be used in large-scale (human)

studies. The aim of the present study was to identify universal biomarkers for CRD

independent of time of day using a transcriptomics approach. Female FVB mice

were exposed to six shifts in a clockwise (CW) and counterclockwise (CCW) CRD

protocol and sacrificed at baseline and after 1 shift, 6 shifts, 5 days recovery and 14

days recovery, respectively. At six time-points during the day, livers were collected

for mRNA microarray analysis. Using a classification approach, we identified a set of

biomarkers able to classify samples into either CRD or non-disrupted based on the

hepatic gene expression. Furthermore, we identified differentially expressed genes

14 days after the last shift compared to baseline for both CRD protocols. Non-

circadian genes differentially expressed upon both CW and CCW protocol were

considered useful, universal markers for CRD. One candidate marker i.e. CD36 was

evaluated in serum samples of the CRD animals versus controls. These biomarkers

might be useful to measure CRD and can be used later on for monitoring the

effectiveness of intervention strategies aiming to prevent or minimize chronic

adverse health effects.


Biomarkers for CRD

61

4

Introduction

Human behavior, physiology and metabolism are subject to daily rhythms,

which are controlled by the circadian clock. This endogenous time keeping

system provides a temporal organization of our body functions in relation to

environmental time and allows us to anticipate to daily recurring events (Mohawk

et al., 2012). Chronic circadian rhythm disturbance (CRD), as encountered by

frequent night shift work or multi time zone travelling might result in an increased

risk for long-term health effects. Indeed, epidemiological studies among shift

workers and flight personnel have associated frequent shift work and jet lag

with an increased incidence of breast cancer, obesity and metabolic syndrome

(Reynolds et al., 2002, Schernhammer et al., 2006, Kubo et al., 2011). These adverse

health effects occur after many years of shift work, and at present it is unclear what

mechanism is causing adverse health effects and how these effects of shift work

can be minimized. The ability to measure chronic CRD associated with shift work

would allow measuring effects of interventions on chronic CRD and monitoring

adversity in shift workers and ultimately will help to design intervention strategies.

Studies on the beneficial effects of interventions to prevent shift work-driven

adverse health outcomes assess effects on CRD using classical circadian markers,

including melatonin, cortisol and body temperature (Mirick and Davis, 2008).

These markers allow monitoring circadian rhythm and acute CRD using multiple

measurements around the clock before health effects occur. In addition to classical

circadian markers, recent research on circadian clock controlled output genes

has shown that up to 10% of the transcribed genes is under circadian control,

providing additional rhythmic markers to estimate body time in blood and tissues

(Panda et al., 2002, Ueda et al., 2004). However, both the classical circadian markers

and cycling clock and clock-controlled gene markers are non-eligible as CRD

markers in large-scale human cohort studies due to two important pitfalls. Firstly,

circadian markers require around the clock measurements, resulting in higher costs

and larger impact on participating subjects compared to single measurements.

Secondly, classical biomarkers are useful for demonstrating acute CRD, but provide

no or only limited information on long-term CRD and accumulation of adversity

over time. To acquire information on biological adversity of CRD and to explore the

effectiveness of CRD preventive measures, new biomarkers are needed to evaluate

the presence of chronic CRD in a time of day independent manner.


Chapter 4

62

4

Shift work involves a multitude of aspects, including phase desynchronization,

light at night, sleep disruption and lifestyle disturbances, all of which potentially

play a role in causing CRD and associated adverse health effects (Fritschi et al.,

2011). Many different shift work schedules are in use, varying in rotation speed

and direction, including forward (counterclockwise) or backward (clockwise)

rotating shift schedules. Experimental studies in which mice were subjected to

(chronic) shifts in the light-dark cycle (as such resembling jet lag), have shown

that both counterclockwise (CCW) and clockwise (CW) schedules cause CRD

(McGowan and Coogan, 2013). Additionally, several human studies have shown

disturbed circadian rhythms by both CCW and CW work schedules, without major

differences between the schedules (Barton et al., 1994, Boquet et al., 2004, Tucker

et al., 2000). However, in aged mice CCW shifts appeared more disruptive than CW

shifts, as evident from the increased mortality (Davidson et al., 2006).

The aim of the present study was to identify universal biomarkers for CRD

independent of rotation direction and time of day. Two different rotations of chronic

jet lag were used to induce CRD. Since blood biomarker discovery is technically

challenging, we selected the liver to identify biomarkers, as the target tissue of

metabolic effects of CRD and as previously used for circadian transcriptomics

studies (Ueda et al., 2004). By comparing the liver transcriptome of animals under

normal, CW rotating and CCW rotating light schedules, we identified a set of

hepatic gene expression markers that report on the presence of CRD. Additionally,

we identified non-circadian, age-independent genes differentially expressed after

CRD compared to baseline that are potentially blood detectable. One candidate

biomarker i.e. CD36 was validated in blood, allowing future use in large-scale

human studies.

Methods

Study designAnimal studies were performed in compliance with national legislation, including

the 1997 Dutch Act on Animal Experimentation, and experiments were approved

by the Animal Experimentation Ethical Committee of the National Institute for

Public Health and the Environment in Bilthoven. All surgery was performed under

isoflurane anesthesia and appropriate analgesics were used to minimize suffering.


Biomarkers for CRD

63

4

Female FVB mice, 8 to 12 weeks of age, were kept under a normal 12:12 hour

light-dark (LD) cycle for approximately three weeks, with Zeitgeber Time 0 (ZT0)

corresponding to lights on. Animals were group-housed and food and water

were provided ad libitum. Prior to the first shift, animals (n=24) were sacrificed

around the clock at four hour intervals (n=4 at each time point). Subsequently,

the remaining group of animals underwent six shifts in either a clockwise (CW) or

counterclockwise (CCW) rotating light schedule. Specifically, mice were exposed

to a phase delay or phase advance of eight hours every five days, respectively (S1

Figure). Hereafter, mice were again kept under LD, referred to as recovery. After

one shift, six shifts, five days recovery and fourteen days recovery n=30 animals

per group were sacrificed around the clock with four hour intervals (n=5 per time

point) by orbital bleeding under Ketamine/Xylazine anesthesia. Liver and blood

were collected for transcriptomics and serum analyses, respectively. A detailed

overview of the experimental design is depicted in S1 Figure.

Four additional mice per group received a radio transmitter (Physio Tel, TA11

TA-F10; Data Sciences, St. Paul, MN) in the peritoneal cavity to record core body

temperature every ten minutes. Body temperature was recorded throughout the

experiment. Cosine curves were fitted using the R statistical software environment

(www.r-project.org) to determine the phase (i.e., peak time) of activity and body

temperature rhythms.

Microarray analysisRNA was extracted from livers (n=4 per time point at baseline and n= 2 per time

point for the CCW and CW groups) using the miRNeasy Mini Kit (Qiagen Benelux,

Venlo, The Netherlands). RNA concentrations were measured using a NanoDrop

ND-1000 Spectrophotometer (Nanodrop Technologies, Wilmington, DE, USA), and

RNA quality was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies,

Amstelveen, the Netherlands).

RNA was processed for gene expression analysis at the Microarray Department

of the University of Amsterdam, the Netherlands, using methods described in

Pennings et al (Pennings et al., 2011). Experimental samples (each corresponding

to RNA from one individual mouse) were labelled with Cy3 and the common

reference sample (made by pooling equimolar amounts of RNA from experimental

samples) was labelled with Cy5. Samples were hybridized to Nimblegen Mus


Chapter 4

64

4

musculus 12 x 135 k microarrays (Roche NimbleGen, Germany). This type of

microarray contains 44,170 gene probes with three spots per probe. Slides were

scanned with an Agilent G2565CA DNA microarray scanner. Feature extraction

was performed with NimbleScan v2.5 (Roche NimbleGen), resulting in a table

containing individual probe signal intensities for both dyes.

The raw data were subjected to a set of quality control checks to ensure comparable

signal average and distribution. Raw microarray data for gene-coding probes

were normalized in R (www.r-project.org) using a four step approach (Pennings

et al., 2011): (1) natural log-transformation, (2) quantile normalization of all scans,

(3) correcting the sample spot signal for the corresponding reference spot signal

and (4) averaging data from replicate probe spots. The normalized data of 44,170

probes were further analyzed in R and Excel (Microsoft Corporation, USA). Probe

to gene annotation data was downloaded from NCBI. Classification and statistical

analysis was performed on the probe level. Significantly predictive or regulated

probe sets were annotated to the corresponding genes for further biological

interpretation. To this end, when multiple probes corresponding to the same

gene were significant, they were counted as one gene in further analysis; probes

that did not correspond to genes according to the current NCBI database were

excluded from further analysis.

Complete raw and normalized microarray data and their MIAME compliant

metadata have been deposited at GEO (www.ncbi.nlm.nih.gov/geo) under

accession number GSE65346.

Classification approachAfter normalization, a classification approach was applied to identify a set of genes

able to classify samples into either circadian rhythm disrupted (CRD) or non-

disrupted (ND) 14 days after the last shift. Three different algorithms were used:

Random forests (RF) (Breiman, 2001), Support vector machine (SVM) (Rifkin et al.,

2003) and Prediction analysis for Microarrays in R (PAM-R) (Tibshirani et al., 2002) (R

statistical software). To ensure time-of-day independence of the classifier, for each

classification algorithm, a ‘leave-one-time point-out’ cross-validation approach

was used for the classification. Here, time point refers to ZT time point. To this end,

the data were repeatedly split into a training set and a test set, in which the training

set comprised the data for all-but-one time point, and the test set the data for the


Biomarkers for CRD

65

4

remaining time point. The prediction model obtained for the training set was used

to predict the test set data. This approach keeps replicate (time point) samples

together and therefore gives a more reliable estimation of the prediction accuracy

than cross-validation with random sample selection. The prediction accuracy was

calculated as the average over all the test set predictions. As each classifier builds

a different prediction model (potentially using different genes) for each training

set, genes included in the majority of the models for each classifier were taken as

consensus gene set for each type of classifier.

As we aim to find biomarkers which applicability does not depend on a specific

choice of algorithm, only genes present in all three classifiers consensus sets were

considered potentially valuable biomarkers to classify CRD versus non-disrupted

(van der Veen et al., 2013, Vandebriel et al., 2010). To validate prediction accuracy of

the consensus gene set after one shift, six shifts and after 5 days recovery, the 14

days recovery dataset was used as the training set to build the prediction model.

Sequential approachTo identify differentially expressed genes (as compared to baseline) after 14 days

recovery for the CCW and CW protocol separately, we performed a one-way

ANOVA with Qlucore Omics Explorer (Qlucore AB, Lund, Sweden) in which p<0.001

was considered statistically significant. CircWave Batch v5.0 software (Hut, R., www.

euclock.org/results/item/circ-wave-batch.html) was used to analyze circadian

rhythmicity of gene expression. P-Values were false discovery rate (FDR) corrected

(Benjamini and Hochberg, 1995); genes with an FDR <0.05 were considered

rhythmically expressed. Genes that were rhythmically expressed at any time

point during the experiment (baseline, 1 shift, 6 shifts 5 or 14 days recovery) were

excluded as potential biomarker. The GenAge Database of ageing-related genes

(www.genomics.senescence.info) was used to identify (human) age-dependent

genes, which were excluded also.

To determine which candidate biomarker genes are potentially detectable

in human serum or plasma, we determined which genes code for proteins

that are annotated in Gene Ontology as extracellular or in UniProt as secreted.

Additionally, we determined which genes have human equivalents that have

been experimentally detected with high confidence in plasma or serum (Farrah et

al., 2011) or as part of the Human Plasma Proteome Project (Nanjappa et al., 2014).


Chapter 4

66

4

Biomarker serum levelsCorticosterone serum levels were determined using ELISA assays (Yanaihara

Institute Inc. Shizuaka, Japan) and subsequently visualized and analyzed using

GraphPad Prism software version 6.04 for Windows (GraphPad Software, San Diego

California USA). Five outliers (out of 251 samples) were excluded based on Grubbs

analysis (alpha 0.1). Phase and circadian rhythm were analyzed using CircWave

Batch v5.0 software. CD36 was determined in serum with a dedicated ELISA assay

(Abcam, Cambridge, United Kingdom). Differences between CRD exposed groups

and baseline were statistically tested using a two-sided student’s t test. P-values

<0.05 were considered statistically significant.

Results

Classical circadian markersFirst, we evaluated whether the two different schedules, counterclockwise (CCW)

and clockwise (CW), affected circadian rhythm by analyzing classical circadian

markers: core body temperature and corticosterone rhythms. At baseline, animals

showed regular daily body temperature rhythms, with peaks at approximately

ZT12. Core body temperature rhythms re-entrain to the new light-dark cycle after

the first shift and following shifts, for both CCW and CW groups (Fig. 1, panels

A & B, respectively).

As expected, corticosterone serum levels at baseline showed a major peak at ZT12.

Additionally, a minor peak at ZT0 was also observed, corresponding with a small

increase in activity at this time point specific for FVB mice (CircWave tau = 12,

p = 0.0003). Differential effects between the two schedules could be observed

after peak phase analysis of corticosterone rhythms (Fig. 2). After the first shift, the

corticosterone rhythm shows peak levels at ZT16 for the CW group, representing

an incomplete shift (p = 0.005). In the CCW group peak time was at ZT20, indicative

of a lack of phase shift at this time point (p = 0.052). Although there is a tendency

toward circadian rhythm, no significant circadian corticosterone rhythm could be

detected in either group upon six shifts (CCW: p = 0.369, CW: p = 0.700). At 5 days

and 14 days after the last shift, rhythms were detected in both groups. After 14

days recovery these rhythms were less robust for CCW and CW (p = 0.093 and p =

0.105, respectively) compared to 5 days after the last shift (CCW: p = 0.007, CW: p =


Biomarkers for CRD

67

4

0.003). Overall, these results show that circadian rhythms of corticosterone levels

heavily disturbed upon prolonged exposure to CRD, but reappear when animals

are back under normal LD conditions.

Figure 1. Double plots of peak phase of core body temperature rhythms. A. the counterclockwise schedule and B. the clockwise schedule. A cosine function was fitted to determine peak phase (n=4 mice per group). Data are presented as mean peak phase ± sd. Grey areas indicate periods of darkness.

A B


Chapter 4

68

4

Figure 2. Corticosterone serum levels for the counterclockwise (CCW) and clockwise (CW) jet lag protocols. Serum levels were determined at baseline, after 1 shift, 6 shifts, 5 days recovery and 14 days recovery. Data are presented as mean ± SD. For the first shift, the light dark schedule before the shift is depicted in grey. At baseline a significant 12-hr rhythm was detected, due to the major peak at ZT12 and another peak at ZT0. CircWave peak phases were indicated with (ê). Note: as circadian rhythmicity of serum corticosterone levels was lost after 6 shifts, peak phases at the 6th shift are just indicative of the best cosine fit.


Biomarkers for CRD

69

4

In summary, both experimental jet lag schedules, CCW and CW rotation affected

the classical circadian markers, indicating CRD. Minor differences in classical

circadian markers were found between the two schedules directly after a shift.

However, for both schedules the effects were transient, largely recovering within

14 days after the last shift.

Predictive set op hepatic transcriptome markersSince liver is the target tissue of metabolic effects and can be used for future studies,

we performed an around the clock analysis of the liver transcriptome at baseline

and after 14 days recovery. To identify a predictive set of hepatic transcriptome

markers for chronic CRD with time of day-independent expression levels, we

applied a classification approach. Three different classification algorithms were

used: RF, SVM and PAM-R (complete overview in Fig. 3a). Using a ‘leave-one-time

point-out’ cross-validation approach, we identified one consensus set of genes per

algorithm optimally classifying samples as CRD versus non-disrupted, after 14 days

recovery independent of time of day. The SVM approach resulted in a consensus

set of 226 probes, RF in 42 and PAM-R in 46. Only the 18 probes, corresponding

to 15 individual genes, present in all three classifiers were considered potentially

robust biomarkers to classify CRD versus non-disrupted (van der Veen et al., 2013,

Vandebriel et al., 2010) (Table 1, S2 Figure). Based on this consensus gene set

prediction accuracy was achieved ranging from 90 % to 98% (S1 Table) showing

that the set of 15 genes could distinguish CRD-exposed animals from non-

disrupted controls with high accuracy independent of sampling time.


Chapter 4

70

4

Figure 3. Schematic overview of the microarray analyses. A. Overview of the classification approach, resulting in a set of transcriptomics biomarkers including 15 individual genes. B. Outline of the sequential approach of the data-analyses from hepatic gene expression profiles to 9 potential blood-detectable protein biomarkers.

Subsequently, the ability of the consensus gene set to detect CRD after one shift,

6 shifts and 5 days recovery was determined. Detection of acute CRD was limited,

as after the first shift only 29% to 33% of the samples of phase shifted animals

were correctly classified as CRD samples, depending on classification method.

Accumulation of CRD was detected in the samples taken after six shifts, here, 92%

to 88% of the samples were correctly classified as CRD exposed. Samples taken

five days after the last shift were also well classified by the gene set, 75% to 88%

depending on the algorithm (S2 Table). Overall, this hepatic transcriptome marker

set is well able to detect chronic CRD independent of the time point the sample is

collected within 14 days recovery.


Biomarkers for CRD

71

4

Table 1. Potential biomarkers for CRD as identified by classification and sequential approach

Gene symbol Function Consensus gene set (classification approach)

Non-invasive blood detectable marker (sequential approach)

Cyp2c29 Cytochrome P450 epoxygenase x x

Cyp2b10 Cytochrome P450 x

Rbp1 Vitamin A transport x x

Sult2a1 Sulfotransferase x

Cd36 Scavenger receptor x x

Ntrk2 Kinase signaling x x

Tusc3 Tumor suppression x

Armcx3 Tumor suppression x

Gspt2 Cell cycle progression x

Snrpn Transcription x x

Tceal8 Transcription x

Fkbp11 Protein folding x

Orm2 Acute phase plasma protein x

Gm3787 Unknown x

Gm9299 Unknown x

D630033O11Rik Unknown x

Igh-VJ558 Immune x

Srgap3 GTPase activity x

Tram1 Translocation proteins x

Non-invasive biomarkers for CRDAlthough predictive of CRD, use of hepatic gene expression markers is still invasive,

where non-invasive methods are preferable. However, direct biomarker discovery

in blood is technically challenging. To find non-invasive biomarkers eligible to

monitor CRD, we aimed to identify potential blood markers for CRD from the gene

expression dataset. Therefore, we selected genes of which expression (i) increased

or decreased with accumulating CRD exposure and (ii) remained deregulated after

14 days recovery. Furthermore, we excluded genes with circadian expression levels

as these have drawbacks earlier described. Compared to control animals, 339 genes

were differentially expressed in mice exposed to the CCW schedule. For mice in

the CW rotation schedule (p <0.001) 129 were differentially expressed, of which 42

genes were significantly expressed in both groups. Of these 42 genes, only genes

with a fold change (FC) larger than 1.2 or smaller than 0.8 were considered relevant

to increase the probability of detectable differences in blood protein levels. This

resulted in 17 genes all showing an approximate pattern of up or down regulation

with accumulating shifts and remaining differentially expressed after 14 days

recovery compared to baseline.


Chapter 4

72

4

Of these 17 genes, 9 genes were found to encode potentially blood-detectable

biomarkers according to (i) annotation as “secreted” or “extracellular”, and/or (ii)

previous proteome-based experimental detection in human plasma or serum:

Cd36, Ntrk2, Igh-VJ558, Srgap3, Tram1, Snrpn, Rbp1, Cyp2b10 and Cyp2c29 (see Fig. 3b

and S3 Table for the sequential flow of gene selection). A substantial overlap with

the classification consensus gene set was found, 5 genes were identified by both

approaches (Table 1). Four out of the 9 genes showed increasing up regulation

with number of shifts and remained up-regulated during recovery and five genes

showed a similar pattern in the opposite direction (Fig. 4).

Figure 4. Potentially blood-detectable marker for CRD. Fold ratio of selected differentially expressed genes encoding potentially blood-detectable protein biomarkers for CRD. Expression of these genes is up-regulated or down-regulated by both the CCW and the CW schedule and remains up or down regulated during recovery.

Based on expression patterns and the availability of a serum ELISA assay, CD36

was selected for validation in blood. At 14 days after the last shift, CD36 serum

levels showed a significant increase of 18% in animals exposed to CCW shifted

light schedules independent of time of day, compared to animals at baseline (Fig.

5). For the animals exposed to the CW light schedule, an 11% increase was found.

For both shifting schedules, the increase in blood levels was less pronounced than

the increase observed in the gene expression data.


Biomarkers for CRD

73

4

Figure 5. Serum levels of the CD36 protein. Serum levels were determined at baseline and 14 days after the last shift for animals subjected to CCW and CW shifted light schedules. Error bars indicate mean ± sd.

Discussion

Frequent shift work results in a disturbance of the circadian rhythm (CRD) and might

on the long-term result in increased health risk. Epidemiologic studies among shift

workers and flight personnel have shown increased risk of breast cancer, obesity

and metabolic syndrome (Kubo et al., 2011, Reynolds et al., 2002, Schernhammer

et al., 2006). The growing 24/7 economy will only lead to increase in shift work and

consequently will adversely affect health. Ideally, the level of chronic CRD should

be detected before the negative health effects occur. To limit potential health

effects, preventive measures to minimize CRD are an attractive option. For these

purposes, measuring the presence of chronic CRD is of importance, however, tools

to measure chronic CRD are currently lacking. The present study aimed to identify

universal biomarkers for chronic CRD using a transcriptomics approach.

To date, CRD and the beneficial effects of preventive measures on CRD are mainly

determined using classical circadian markers such as melatonin, cortisol or body

temperature (Neil et al., 2014). There is a need of markers that show the cumulative

effects of CRD, since the effects on available markers are transient showing only

acute effects of CRD. Corticosterone rhythms were heavily disturbed after six

subsequent shifts of the light dark schedule and reappear after 5 days recovery,

whereas body temperature remains rhythmic and was found to re-entrain to the

new light dark schedule within 4-5 days, even after multiple shifts. The present


Chapter 4

74

4

study identified chronic, non-transient biomarkers for chronic CRD 14 days after

the last shift, based on hepatic gene expression. The potentially blood detectable

biomarkers report on the presence of chronic CRD, showing an increase or

decrease after multiple shifts and non-reversible deregulation upon recovery. The

classification markers (consensus gene set) could be used directly after multiple

shifts and 5 or 14 days after the last shift, which makes good candidates to

determine acute shift work related CRD.

Ideally, for large-scale molecular epidemiology and experimental studies

biomarkers of CRD are time-independent. The current available circadian markers

require 24-hr measurements, which is challenging in both experimental and

field studies. Recent attempts have been made to identify transcriptomics- and

metabolomics-based time of day independent biomarkers. In these studies, jet

lag and clock mutant mice were used (Minami et al., 2009, Ueda et al., 2004),

allowing detection of acute and transient effects of jet lag but not chronic

effects. It is challenging to model human shift work since it involves a multitude

of aspects, including phase desynchronization, changed social patterns, activity,

sleep, nutrition light exposure and sun exposure. Using a chronic jet lag model in

mice, and as such mainly mimicking phase desynchronization, we were able to

identify biomarkers detecting chronic CRD that can be studied in a single sample,

collection of which is independent of time of day. Further studies should point out

whether other aspects of shift work (e.g. changes in activity, light-at-night, altered

nutritional timing) will affect the same genes.

The set of liver-transcriptome based biomarkers includes genes with a variety of

functions. The current study suggests that CD36 could potentially be a blood-

detectable biomarker for CRD. CD36 is a scavenger receptor present on many

mammalian cell types with a broad range of cellular functions (Febbraio et al.,

2001). It has been suggested that CD36 in plasma might represent a marker of the

metabolic syndrome (Handberg et al., 2006), a condition that was found associated

with frequent shift work (Wang et al., 2014). Furthermore, CD36 was shown to play

an important role in breast tumorigenesis (DeFilippis et al., 2012, Uray et al., 2004)

potentially associated with the observed increase in breast cancer risk found in

shift workers (Wang et al., 2013). Although CD36 is interesting as CRD biomarker,

other potentially blood-detectable biomarkers should not be neglected.

Especially, the genes that were found upregulated upon CRD are of interest for


Biomarkers for CRD

75

4

further investigation. For example, Ntrk2 shows a similar induction pattern as CD36

and has also been shown to play a role in breast cancer cell survival and obesity

(Xu et al., 2003, Vanhecke et al., 2011). In depth analysis of biological relevant data

in this study is limited due to our aim to select non-circadian biomarkers, for which

circadian genes were excluded. Full analysis of biological processes affected by

CRD requires inclusion of these circadian genes and is subject for further studies.

Our study represents two approaches to identify the most valuable CRD markers

based on hepatic gene expression; firstly, an optimal CRD classification set and

secondly, a selection of potentially blood detectable biomarkers. An important

step that needs to be taken before the identified biomarkers can be applied in

experimental and large-scale cohort studies is validation for CRD in humans. For

blood-detectable markers, the challenge of the transcriptomics approach is the

translation of hepatic gene expression to protein levels in blood. We found that in

our homogenous mouse model, CD36 serum levels were also increased in animals

exposed to CCW and CW shifted light schedules; however, the increase was

smaller compared to gene expression. Potentially, serum CD36 originates from

other sources than liver alone, since CD36 is present in many mammalian cell types

(Febbraio et al., 2001). Another part of the validation process is the exclusion of post-

translational rhythmicity, which is not precluded by the lack of a transcriptional

rhythm. In human samples, heterogeneity and variation between samples is much

larger and small increases in biomarker blood levels may remain undetected in

cross-sectional studies. Preferably, to obtain less inter-individual variation one

would opt for longitudinal measurements including baseline measurements per

individual before commencing shift work rotations. Furthermore, anchoring with

phenotypic endpoint is required to use the selected biomarkers for intervention

studies without the need for long-term end-point studies. For this, it should be

noted that previous studies have shown that comparable chronic CRD protocols

increased negative health effects in mice (Davidson et al., 2006, Filipski et al., 2004).

In conclusion, our study identified a chronic CRD gene-set, comprising 15 genes,

potentially useful to study CRD induction by different aspects of shift work and

reduction by interventions. Furthermore, we identified 9 candidate genes for

blood-detectable biomarkers of CRD, including CD36. Upon validation, these

biomarkers provide valuable tools for evaluating CRD in both experimental animal

and human studies set up to identify preventive measures for adverse health

effects.


CHAPTER 5

Diurnal variation of hormonal and

lipid biomarkers in a molecular epidemiology-

like setting

Adapted from:

Diurnal variation of hormonal and lipid biomarkers in a molecular

epidemiology-like setting

Linda W.M. van Kerkhof*, Kirsten C.G. Van Dycke*, Eugene H.J.M. Jansen,

Piet K. Beekhof, Conny T.M. van Oostrom, Tatjana Ruskovska, Nevenka Velickova,

Nikola Kamcev, Jeroen L.A. Pennings, Harry van Steeg, Wendy Rodenburg


PLoS One, 2015, 18;10(8)


Chapter 5

78

5

Abstract

IntroductionMany molecular epidemiology studies focusing on high prevalent diseases, such

as metabolic disorders and cancer, investigate metabolic and hormonal markers.

In general, sampling for these markers can occur at any time-point during the day

or after an overnight fast. However, environmental factors, such as light exposure

and food intake might affect the levels of these markers, since they provide input

for the internal time-keeping system. When diurnal variation is larger than the

inter-individual variation, time of day should be taken into account. Importantly,

heterogeneity in diurnal variation and disturbance of circadian rhythms among

a study population might increasingly occur as a result of our increasing 24/7

economy and related variation in exposure to environmental factors (such as light

and food).

AimThe aim of the present study was to determine whether a set of often used

biomarkers shows diurnal variation in a setting resembling large molecular

epidemiology studies, i.e. non-fasted and limited control possibilities for other

environmental influences.

Results

We show that markers for which diurnal variation is not an issue are

adrenocorticotropic hormone, follicle stimulating hormone, estradiol and high-

density lipoprotein. For all other tested markers diurnal variation was observed

in at least one gender (cholesterol, cortisol, dehydroepiandrosterone sulfate, free

fatty acids, low-density lipoprotein, luteinizing hormone, prolactin, progesterone,

testosterone, triglycerides, total triiodothyronine and thyroid-stimulating

hormone) or could not reliably be detected (human growth hormone).

Discussion: Thus, studies investigating these markers should take diurnal variation

into account, for which we provide some options. Furthermore, our study indicates

the need for investigating diurnal variation (in literature or experimentally)

before setting up studies measuring markers in routine and controlled settings,

especially since time – of- day likely matters for many more markers than the ones

investigated in the present study.


Diurnal variation in hormonal and lipid biomarkers

79

5

Introduction

Many molecular epidemiology studies focusing on high prevalent diseases, such

as metabolic disorders and cancer (Danaei et al., 2011, WHO, 2013), make use of

metabolic and hormonal markers (for examples see references (Hadji et al., 2014,

Jiang et al., 2014). Metabolic markers might serve as important (early) indicators

for metabolic disorders, including cardiovascular diseases and type 2 diabetes, and

hormonal disbalance is often studied in large epidemiological settings since these

are associated with high incidence cancers, such as breast cancer (Mester et al., 2010,

Stevens, 2005, Ferlay et al., 2015). In general, for these markers sampling can occur

at any time-point during the day or after an overnight fast. However, environmental

factors, such as light exposure and food intake might affect the levels of these

markers, since they provide input for the internal time-keeping system (Carneiro

and Araujo, 2012, Dibner et al., 2010, Hastings et al., 1998, Mohawk et al., 2012).

Several markers are well known for their diurnal variation, for example cortisol and

melatonin (Gamble et al., 2014). Importantly, heterogeneity in diurnal variation and

disturbance of circadian rhythms among a study population might increasingly

occur as a result of our increasing 24/7 economy and related variation in exposure

to environmental factors (such as light and food). For example, previous studies

have indicated that there are substantial changes in the blood transcriptome after

experiencing insufficient or mistimed sleep (Archer et al., 2014b, Moller-Levet

et al., 2013). This might render accurate determination of markers in molecular

epidemiology studies challenging. When the diurnal variation is larger than the

inter-individual variation, time of day should be taken into account in molecular

epidemiology studies. In this study, we determine whether a set of biomarkers

relevant for metabolic disorders and hormone-associated cancers, consisting of

endocrine and sex hormones and lipids, shows diurnal variation.

The diurnal variation of various blood markers has previously been studied in

different settings, ranging from a relatively uncontrolled routine setting to very

tightly controlled laboratory settings. Most information on diurnal variation is

derived from these controlled laboratory studies in which factors influencing

diurnal variation (such as food intake and/or sleep/wake cycle) are controlled

(for example see references (Klingman et al., 2011, Li et al., 2013, Jung et al., 2010,

James et al., 2007, Davies et al., 2014, Ang et al., 2012, Dallmann et al., 2012)). Due

to the large-scale set up of some molecular epidemiology studies, the possibility


Chapter 5

80

5

for standardization of sample collection to time of day and food consumption is

limited. Therefore, most molecular epidemiology studies collect blood samples to

study biomarkers for adverse health outcomes without standardization for time-

point, food intake or sleep.

The aim of the present study was to investigate diurnal variation of a set of

markers (relevant for metabolic disorders and hormone-associated cancers) in a

routine setting in males and females, resembling molecular epidemiology studies:

namely, non-fasted and limited control for other environmental influences. In

addition, classical markers to study circadian rhythms in chronobiology research

are measured: cortisol and components of the molecular biological clock i.e. ‘clock

genes’. We determined cortisol levels and clock gene expression levels in blood

throughout the day to study diurnal variation of these classical markers in a routine

setting.

Methods

Study design and ethics statementApproval for all procedures was obtained from the ethical committee of the Goce

Delchev University in Stip, Republic of Macedonia. All participants signed an

informed written consent. This was provided in Macedonian and the translation

would be as follows: “I confirm that I am informed about the goals of the Circadian

study and voluntary participate in it. I also confirm that all information that I provided

with regards to my general health condition and lifestyle is correct”. Samples of 17

healthy volunteers, 10 women and 7 men were analyzed in this study. Inclusion

criteria were: -age over 21 years old, - apparently healthy without any acute or

chronic disease (general medical examination without laboratory analyses), - not

taking any drugs, multivitamins and/or supplements, -non-smoking, - not more

than 1 (for women) or 2 (for men) alcoholic beverages per day, - no shiftwork and

have a normal circadian rhythm (sleeping at nighttime).

The mean age of the males was 24.9 ± 7.2 years and 21.7 ± 0.5 years for the female

volunteers. The mean weight was 84.1 ±13.9 kg for the males and 58.5 ±13.6 kg

for the females. For more participants characteristics and individual data, see S1

Table. The samples were collected around the clock at four-hour intervals. The first



81

5

sample was collected at 08:00 AM after overnight fasting. Participants were asked

when they slept between the first and last sampling times (8:00 - 4:00 AM). Total

sleep duration between these time-points was 3.4 ± 0.9 hr for males and 3.2 ±1.9

for females. For individual sleeping duration and timing of sleep episodes, see

S1 Table. Volunteers were asked to consume their meals one hour after sample

collection and otherwise continue their normal daily routines. Meal timing and

composition was not further restricted or standardized and meals were consumed

at home. EDTA plasma and buffy coat samples (2.7 mL) and serum samples (7.5

mL) were collected in a routine clinical chemistry laboratory by venipuncture

of the antecubital vein, using the commercial blood sampling method Sarstedt

(Sarstedt AG & Co., Nümbrecht, Germany). Individuals were allowed to leave the

laboratory after sample collection and thus were required to wake up and come

to the laboratory for sample collection at night.

Plasma and serum analysesTwo methods were used to determine hormonal and lipid biomarkers, an

overview can be found in S2 Table. Free fatty acids (FFA), high-density lipoprotein

(HDL), low-density lipoprotein (LDL), triglycerides (TG) and cholesterol (CHOL)

were determined with an auto-analyzer (Unicel DxC 800, Beckman-Coulter,

Woerden, the Netherlands), using kits from Beckman-Coulter. Cortisol (CORT),

dehydroepiandrosterone sulfate (DHEAS), estradiol (E2), human growth

hormone (hGH), prolactin (PRL), progesterone (PRG), testosterone (TEST), total

triiodothyronine (TotT3), thyroid-stimulating hormone (TSH) were determined

with an immune-analyzer (Access-2, Beckman-Coulter) using dedicated kits

from Beckman-Coulter. Levels of adrenocorticotropic hormone (ACTH), follicle

stimulating hormone (FSH), and luteinizing hormone (LH) were determined using

commercially available Milliplex Kits (Millipore Corporation, Billerica, MA, USA). All

measurements were performed on the same day. The intra-assay variation was

determined with three quality control samples (8 tests) for markers measured

on the auto-analyzer and with two quality control samples (5 tests) for markers

measured on the immune-analyzer. The CVs of markers measured with the

Luminex technique were obtained from the manufacturer. See S2 Table for all CVs.

Blood RNA analysisExpression levels of clock genes BMAL1 and PER1 were determined in buffy coats

of a subset of volunteers (n=6, female), using quantitative reverse transcription


Chapter 5

82

5

polymerase chain reaction (RT-PCR). All oligonucleotide primers were obtained

from Life Technologies (Bleiswijk, The Netherlands). Total RNA was extracted from

buffy coats stabilized with Qiazol, using the miRNeasy Mini Kit (both Qiagen

Benelux, Venlo, The Netherlands). cDNA was made using the high capacity cDNA

reverse transcription kit (Applied Biosystems by Life Technologies, Bleiswijk,

The Netherlands). qPCR was performed on the 7500 Fast Real-Time PCR System

(Applied Biosystems by Life Technologies, Bleiswijk, The Netherlands) using

Taqman Fast Universal PCR Master Mix and Taqman Gene Expression assays for

BMAL1 (Hs00154147_m1) and PER1 (Hs01092603_m1) all according to protocol of

manufacturer (Applied Biosystems by Life Technologies, Bleiswijk, The Netherlands).

StatisticsTo detect all types of diurnal variation, parameters were tested using two types

of analyses: cosine fitting analysis and Repeated Measures ANOVA (RM-ANOVA).

Depending on the shape of the day curves, these analyses can be overlapping

and/or complementary. The RM-ANOVA detects effects of time, i.e. are the levels

of the marker different among time-points. The cosine analysis enables detection

of rhythms that follow a cosine curve, in manner that is occasionally more sensitive

than the RM-ANOVA. For the cosine analysis, CircWave Batch v5.0 software (Roelof

Hut, www.euclock.org) was used. A closer fit to cosine gives lower p-value, p <

0.05 is assumed a significant cosine curve fit. . Repeated Measures ANOVA (RM-

ANOVA) was performed using GraphPad Prism software version 6.04 for Windows

(GraphPad Software, San Diego California USA, www.graphpad.com). The

Greenhouse-Geisser epsilon indicated that sphericity could not be assumed in

most analyses, therefore, Greenhouse- Geisser correction was applied. For the RM-

ANOVA, p < 0.05 was considered statistically significant. If CircWave analysis, RM-

ANOVA, or both indicated that signification variation was present, we identified

the markers as having diurnal variation in our experiment. Gene expression of PER1

and BMAL1 was given as relative expression compared to the mean.

Considering the known difference in levels of several makers between genders,

all statistical analyses were performed separately for males and females. The aim

of this study was to identify markers with diurnal variation, therefore, differences

between males and females were not statistically tested. For all raw data, see S1

File.



83

5

Results

We investigated diurnal variation of a set of markers (relevant for metabolic

disorders and hormonal cancers) in a routine setting in males and females. To

determine the diurnal variation we performed two analyses: 1) Cosine fitting

analysis: using CircWave software we determined whether the level of the marker

follows a cosine rhythm; 2) RM-ANOVA analysis: determine effects of time of day

on parameter levels. If analysis 1, analysis 2 or both indicated that significant

variation was present we identified the markers as having diurnal variation in our

experiment.

HormonesCosine analysis showed that for two of the eleven investigated hormonal markers,

a significant cosine curve fit was present in both genders: thyroid-stimulating

hormone (TSH) and total triiodothyronine (totT3) (Fig.1). The levels of TSH are

highest during the middle of the night, where in males (m) the peak is observed

slightly earlier (Clock Time (CT) = 01:54, p = 0.001, amplitude as percentage of the

median = 75.89%) compared to females (f ) (CT = 03:32, p = 0.006, amplitude =

74.26%). Levels of totT3 are highest at the beginning of the day (m: CT = 06:18, p

0.025, amplitude = 10.14%; f: CT = 07:04, p = 0.041, amplitude = 8.08%). Progesterone

(PRG) had a significant cosine curve fit in males, but not in females (m: CT = 07:52,

p = 0.003, amplitude = 117.78%; f: p = 0.870) (Fig. 1). For testosterone (TEST), a

trend towards a cosine curve fit was present in males, but not females (m: CT =

07:33, p = 0.055, amplitude = 38.38%; f: p = 0.282) (Fig. 1). For a complete overview

of CircWave analyses for all 11 hormonal markers, see Fig 1-3 and S3 Table.

For the two markers with a significant cosine fit in both genders, TSH and TotT3,

an effect of time was observed with RM-ANOVA as well (for statistics see S3 Table),

although for TotT3 in females only a trend is observed (p = 0.078). In addition, for

four markers where a cosine curve could not significantly be fitted, a significant

effect of time was observed, for TEST, DHEAS, PRG and prolactin (PRL) (Fig. 1 + 2).

RM-ANOVA shows that TEST and DHEAS are present at different levels during the 6

time-points in both genders (TEST: m: p 0.021, f: p = 0.023; DHEAS: m: p = 0.045, f: p

= 0.002). Furthermore, PRL and PRG are significantly different over the time-points

in one gender: PRL (m: p = 0.014, f: p = 0.118) and PRG (m: p = 0.020, f: p = 0.272).

For complete overview of the results of the RM-ANOVA on all time-points, see

Fig 1-3 and S3 Table.


Chapter 5

84

5

Figure 1. Levels of TSH, TotT3, PROG, and TEST during the day, measured at four-hour intervals for male (left panels) and female (right panels) volunteers. Time indicates clock time. Values at 8-hr were double plotted to help visualize daily patterns. Data represent mean ±sd.



85

5

Figure 2. Levels of DHEAS, PRL, LH, and FSH during the day, measured at four-hour intervals for male (left panels) and female (right panels) volunteers. Time indicates clock time. Values at 8-hr were double plotted to help visualize daily patterns. Data represent mean ±sd.


Chapter 5

86

5Figure 3. Levels of ACTH, hGH, and E2 during the day, measured at four-hour intervals for male (left panels) and female (right panels) volunteers. Time indicates clock time. Values at 8-hr were double plotted to help visualize daily patterns. Data represent mean ± sd.



87

5

Figure 4. Levels of different lipids during the day, measured at four-hour intervals for male (left panels) and female (right panels) volunteers. Time indicates clock time. Values at 8-hr were double plotted to help visualize daily patterns. Data represent mean ±sd.


Chapter 5

88

5

LipidsFive types of lipids were examined; free fatty acids (FFA), triglycerides (TG), total

cholesterol (CHOL), low-density lipoprotein (LDL), and high-density lipoprotein

(HDL) (Fig. 4). CircWave analysis indicated that FFA levels follow a cosine curve in

males (p = 0.012, amplitude 74.91%), but not significantly in females (p = 0.359,

amplitude = 80.70%). The levels of FFA were highest at the beginning of the active

period in males (CT = 08:31). A trend towards a cosine curve fit was observed

for TG in females (p = 0.050, amplitude = 50.33%), but not in males (p = 0.105,

amplitude = 70.36%) (Fig. 4). Visual inspection of the levels of FFA and TG in Fig. 4

implies that the differences between males and females for these markers is likely

related to the variation observed within the groups, since the highest levels are

observed at similar time-points for both genders. For complete overview of all

markers, see Fig. 4 and S4 Table.

Despite the limited presence of significant cosine curve fits observed among the

different lipid markers, for several markers RM-ANOVA shows significant effects of

time: for CHOL and LDL in both genders (CHOL: m: p = 0.003, f: p = 0.003; LDL: m:

p = 0.002, f: p = 0.006), for TG in males and a trend was present in females (m: p

= 0.048, f: p = 0.063) (S4 Table). The levels of FFA show a trend towards an effect

of time in males (m: p = 0.069, f: p = 0.221). For HDL, no significant effect was

observed in any gender (m: p = 0.110, f: p = 0.213). The highest level of most lipids

including FFA, LDL, and CHOL was observed at the beginning of the active phase,

the highest level of TG in males was observed during the middle of the day (Fig. 4).

In summary, for all lipid markers, except HDL significant or trends towards diurnal

variation are observed.

Classical circadian markers

The levels of cortisol were investigated in all participants (n=17), clock genes

PER1 and BMAL1 were investigated in a subset (n=6, females) to gain insight in

classical circadian markers in uncontrolled conditions (Fig. 5). Cortisol levels show

a significant cosine curve fit, are highest at the beginning of the day (m: CT =

07:47; f: CT = 07:29), and the rhythm has a relative large amplitude (m: p < 0.0005,

amplitude = 129.11%; f: p < 0.0005, amplitude = 161.63%) (Fig. 5, upper panels).

The levels of PER1 and BMAL1 mRNA show a significant cosine curve fit as well

(PER1 p = 0.002, amplitude = 80.15%; BMAL1 p = 0.035, amplitude = 35.55%). Levels



89

5

of PER1 were highest during the morning (CT = 8:07), while levels of BMAL1 were

highest during the afternoon (CT = 14:12). For complete overview of statistics see

S5 Table.

Figure 5. Around the clock levels for classical circadian markers cortisol, PER1 and BMAL1. Upper panels show cortisol levels in male (left panels) and female (right panels) volunteers, determined in serum. mRNA expression of BMAL1 and PER1 in buffy coats are depicted in the lower panels (females subset n=6). All were measured at 4 hour intervals, where time indicates clock time. Values at 8-hr were double plotted to help visualize daily patterns. Data represent mean ±sd.

Discussion

The aim of the present study was to investigate diurnal variation in a set of

hormones, metabolic markers, and classical circadian markers in serum, plasma

and peripheral blood mononuclear cells (PBMCs), in a routine non-fasted molecular

epidemiology-like setting, with no restrictions on sleep or diet.

HormonesMarkers without diurnal variation

For markers without diurnal variation, taking the time-point of sampling into

account is not required. This study observed no diurnal variation or trends towards


Chapter 5

90

5

diurnal variation in any gender in four of the eleven hormonal markers investigated,

ACTH, hGH, E2, and FSH. Previous published studies performed in a different type

of setting than the present study, namely a controlled (laboratory) settings do show

diurnal variation in ACTH levels in males , with the highest level being observed at

CT 4.00 h (Li et al., 2013) or CT 7.00 h (Nicolau and Haus, 1989), whereas no diurnal

was observed in females (Nicolau and Haus, 1989). Interestingly, in the present

study, levels were also highest at CT 4.00, indicating similar patterns. However,

in our study larger amount of variation is expected than a controlled laboratory

setting potentially causing the diurnal variation of ACTH undetectable within the

inter-individual variations. Since during day time-points, when sampling is usually

performed in a cohort study, variation in ACTH levels is minimal. Thus, diurnal

variation seems not to be a major issue for ACTH in a molecular epidemiology

setting. However, considering the previous studies on ACTH and that in a large

cohort setting power will be increased, diurnal variation might become larger than

the inter-individual variation and caution is required. In addition, if subjects are

investigated which possibly have a circadian disturbance (e.g. night shift workers),

variation in levels of the markers during night time-points should be taken into

account as well (see recommendations section for suggestions).

For hGH, former reports in controlled settings have shown diurnal variation in

males (Mazzoccoli et al., 2011, Nicolau and Haus, 1989). However, it is important

to note that it is known that hGH is released in pulses with levels varying up to 4

times (Riedel et al., 1995). Hence, in the present study with sampling intervals of 4

hours these pulse releases might interfere with the detection of diurnal variation.

Indeed, in the present study at several time-points large variations in the levels of

hGH are observed, particularly in males. This indicates that hGH is a difficult marker

to measure in a non-time-controlled cohort setting, since possible differences are

easily masked by the large amount of variation present for this marker.

For estradiol, the absence of diurnal variation in the present study is in line

with previous studies that reported the absence of diurnal variation in females

(Brambilla et al., 2009, Mortola et al., 1992), although diurnal variation is observed

in estradiol levels in elderly females (mean age 77 years) (Nicolau and Haus, 1989).

Together, these data indicate that in a cohort study with non-elderly adults, diurnal

variation of estradiol is not an issue.



91

5

For FSH, conflicting data have been published previously. In females, the absence

of a diurnal variation was reported in studies using (mildly) controlled settings

(Klingman et al., 2011, Nicolau and Haus, 1989), while a third study reported

diurnal variation in females during different phases of the menstrual cycle, with

the highest levels of FSH being observed during the afternoon (Mortola et al.,

1992). For males, conflicting results have been reported as well: a study by Spratt

et al. showing the absence of diurnal variation in males (Spratt et al., 1988), while

Nicolau et al., reported the presence of a diurnal variation in elderly males (Nicolau

and Haus, 1989). Possibly, age has a role in the diurnal variation of FSH. However,

the present study indicates that in a routine setting, time of day is not a major

source of variation for FSH levels.

Markers with diurnal variation

For eight of the twelve investigated hormonal markers, diurnal variation was

observed in at least one gender. Markers with the most robust diurnal variation in

both genders were TSH and T3. This is in line with previous findings for these markers

(Hirschfeld et al., 1996, Klingman et al., 2011, Mazzoccoli et al., 2011, Nicolau and

Haus, 1989). For DHEAS, a previous study reported circadian rhythmicity in males

and females (Nicolau and Haus, 1989), however, in the present study the rhythm of

DHEAS could not be fitted to a cosine curve. The levels of DHEAS differed among

the time-points and the pattern of expression (e.g. highest levels in the afternoon)

is comparable to the previously published results, confirming diurnal variation of

DHEAS.

For testosterone, diurnal variation was not robust, but significant changes were

observed when the time-points were compared in both genders. Diurnal variation

in testosterone is in line with previous findings (Diver et al., 2003, Nicolau and Haus,

1989, Ostrowska et al., 1998, Plymate et al., 1989, Spratt et al., 1988, Brambilla et al.,

2009). The study by Spratt et al. showed that the robustness of the diurnal variation

of testosterone was largely dependent on the frequency of sampling.

For several of the other markers, the robustness of diurnal variation varied among

genders. Diurnal variation was often more pronounced in males. Since in the

present study menstrual period was not assessed, nor was the use of contraceptives,

this might explain the less robust results in females. For example, progesterone

showed diurnal variation in males, but not females. For females, it has been shown


Chapter 5

92

5

that the rhythmicity of progesterone varies with the menstrual period, including

the timing of the highest levels (Fujimoto et al., 1990, Nicolau and Haus, 1989,

Veldhuis et al., 1988). Interestingly, 5 females show high progesterone levels,

which are indicative of the luteal phase. Of these females, two show a significant

or trend (p-value between 0.05 and 0.1) towards an individual cosine curve fit

(#6: p=0.0059 and #7: p=0.05222). These results suggest that indeed variation in

menstrual period, and possibly contraceptives use, interfere with measurements

of progesterone in a cohort setting.

For the final two markers, LH and PRL, only mild amounts of diurnal variation

were observed. For LH previous studies in females have shown that rhythmicity

is dependent on the menstrual cycle and that peak timing differs among women

(Klingman et al., 2011, Mortola et al., 1992, Veldhuis et al., 1988), which likely

explains the results of the present study. For prolactin, in the present study only

a trend towards a cosine curve fit was observed in females and effects of time

(by RM-ANOVA) in males. This is in line with previous studies that have shown

diurnal variation of prolactin in females in controlled settings (Birketvedt et al.,

2012, Fujimoto et al., 1990), and higher levels of prolactin during the night in males

(Miyatake et al., 1980). Together, these results indicate that for prolactin time of

day should be taken into account (see recommendations section for suggestions).

LipidsOf the 5 lipid markers, only HDL showed no diurnal variation (no cosine curve fit

and no effect of time by RM-ANOVA). Previously, it has been shown that a large

subset of lipids shows diurnal variation (for examples see references: Miida et al.,

2002, Chua et al., 2013, for review see reference: Gooley and Chua, 2014), however,

the rhythmicity of these markers is highly variable among individuals (Chua et

al., 2013). Furthermore, it is known that many lipids are directly regulated by the

endogenous circadian clock (Dallmann et al., 2012, Ang et al., 2012, Gooley and

Chua, 2014), lipid metabolism is related to sleep (Davies et al., 2014, Ollila et al.,

2012), and that levels of FFA, TG and LDL are dependent on meal-timing (Yoshino

et al., 2014, Niklowitz et al., 2006) The results of the present study, diurnal variation

in LDL, TG, FFA and total cholesterol are thus in line with previous findings. These

results show that for these markers time of day needs to be taken into account (see

recommendations section below for suggestions). Previously, diurnal variation has

been observed for HDL in a controlled setting (Miida et al., 2002), however, our



93

5

study indicates that this variation does not exceed the inter-individual variation in

a routine setting.

Classical circadian markersIn chronobiology research, several markers are often used to study circadian

rhythm. Hence, these can be named as ‘classical circadian markers’. One of these

markers is cortisol, which is often used to study circadian rhythm in humans. In

addition, components of the molecular biological clock can be used to study

circadian rhythm, i.e. ‘clock genes’. Since these genes are part of the core clock

mechanism in cells, their rhythm reflects the circadian rhythm in these cells. We

show that the ‘classical circadian markers’ investigated in this study (cortisol and

the clock genes PER1 and BMAL1) have a robust cosine curve fit in our study

population, which resembles a routine non-fasted molecular epidemiology

setting. Our findings for cortisol (rhythmic expression with high levels during the

early morning), are in line with previous literature. Multiple studies, in a range of

different settings, have reported similar findings (for review see reference (Gamble

et al., 2014), for examples see references (Kusanagi et al., 2008, Takimoto et al., 2005,

Teboul et al., 2005)).

In our study PER1 expression fits to a cosine curve with the highest levels in the

morning, which is also consistent with previous studies. For example, James et

al. observed rhythmic PER1 expression with peak levels on average 2:36 h after

waking (±1:47 h) in PBMCs (James et al., 2007), Kusanagi et al. observed rhythmic

PER1expression with peak expression on average at 7:42 h in PMBCs (Kusanagi et

al., 2008), and Takimoto et al. reported peak levels at 06:00 h in whole blood cells

(Takimoto et al., 2005). For BMAL1, previous studies have indicated that rhythmicity

of BMAL1 gene expression in PBMCs is highly variable amongst individuals (Teboul

et al., 2005, James et al., 2007). For example, BMAL1 expression has been reported

to be high during the middle of the day (comparable to our study) (James et

al., 2007, Teboul et al., 2005), during the night (Takimoto et al., 2005), during the

evening (Teboul et al., 2005), or not being rhythmic (Kusanagi et al., 2008). Teboul

et al. suggested that this variation might be related to chronotype, since in their

study two groups of individuals could be distinguished: peak expression of BMAL1

during the middle of the day (12:30 hr) or during the evening (21:45 hr).

Together, our results shows that thediurnal variation, including the acrophase of

the fitted cosine curves, of several classical markers in our study is comparable to


Chapter 5

94

previous studies using (mildly) controlled settings (James et al., 2007, Kusanagi et

al., 2008, Takimoto et al., 2005, Teboul et al., 2005).

Concluding remarks and reccomendationsThe present study showed that markers for which diurnal variation (circadian

rhythms and/or time-of-day-effects) is not an issue in a routine setting are ACTH,

FSH, E2 and HDL. For all other tested markers diurnal variation was observed, which

exceeded the inter-individual variation in a routine setting. It is important to note

that we tested this in a small test group (n=17). In a large cohort with numerous

subjects, levels of inter-individual variation might decrease due to larger sample

sizes, making the contribution of diurnal variation relatively larger. Hence, for some

of the markers without detectable diurnal variation in the present study, diurnal

variation might play a role in larger studies. Nevertheless, our study provides an

indication for which markers diurnal variation needs to be taken into account in a

routine setting.

Taking diurnal variation into account can be done in several ways depending on

the study design and marker characteristics (such as dependent on food intake,

sleep/wake cycle etc.). For example, measurements can be taken at a single time-

point or if sampling at a single time-points is not possible, registration of sampling

time and their relevant time-points (e.g. timing of food intake) is an alternative

method to take into account diurnal variation. Apart from a research setting,

several of the markers in our study are measured in a clinical setting as well. In

general, a substantial amount of variance is incorporated in reference levels for

diagnostic purposes. However, for biomarkers with large diurnal variation it might

be beneficial to take multiple samples during the day or define reference levels

that take into account time-of-day.

Our finding that most markers tested show diurnal variation is in line with recent

studies showing that many genes are regulated by the circadian clock. For example,

a recent study has shown that 43% of the protein coding genes shows diurnal

variation in at least one organ in mice (Zhang et al., 2014a), as a consequence

one could expect diurnal variation in many serum or plasma protein levels as well.

This large set of markers with diurnal variation indicates the need for investigating

diurnal variation before setting up studies measuring markers in routine and

controlled settings, since time- of- day likely matters for many more markers than

the ones investigated in the present study.


CHAPTER 6

Attenuation of circadian rhythmicity in hepatic gene expression upon

chronic alternating light cycle exposure

Kirsten C.G. Van Dycke, Jeroen L.A. Pennings, Conny T.M. van Oostrom,

Linda W.M. van Kerkhof, Gijsbertus T.J. van der Horst, Harry van Steeg,

Wendy Rodenburg

In preparation


Chapter 6

98

Abstract

Disturbance of circadian rhythms in shift workers results in increased health

risks, including obesity and cancer. Previously, we have shown that mice gain

more weight when chronically exposed to conditions that cause circadian

rhythm disturbance (CRD). The aim of the present study was to understand the

mechanism(s) underlying the observed overweight phenotype in mice. As the liver

is a key metabolic organ and regulator of energy metabolism, liver transcriptome

analysis enables the identification of disturbed processes involved in the obese

phenotype. Mice were exposed to a weekly alternating light-dark schedule for

18 weeks and 7 days after the last shift gene expression was analyzed over a 24-

hour period to allow identification of more permanent affected processes upon

long-term CRD. Overall, the effects on gene expression were minor compared to

previous studies where the transcriptome was studied directly after impingement

on the circadian system. However, an overall attenuation of circadian rhythmicity

was found in hepatic gene expression, associated with a wide variety of processes,

including cell cycle control and DNA damage processes. The largest decrease in

amplitude was found in genes involved in the balance between lipid and glucose

metabolism: Ppard, Pdk4, Slc16a5 and Igfbp-1. In conclusion, metabolic flexibility

might be compromised by long-term exposure to CRD, resulting in increased

weight gain in mice. Further research should elucidate the relevance of acute

versus more permanent disturbed processes for adverse health outcomes to

develop evidence based preventive measures.


Attenuation of circadian gene expression upon CRD

99

6

Introduction

The circadian system imposes 24 hour rhythmicity on behavior, physiology and

metabolism and allows us to anticipate daily recurring changes in our environment.

This internal timekeeping system is composed of a light-entrainable central clock

in the suprachiasmatic nucleus (SCN) located in the brain, and peripheral clocks in

all other tissues in the body. On a molecular level, circadian rhythms are generated

by a set of clock genes that act in an auto-regulatory transcriptional-translation

feedback loop (TTFL) and confer rhythmicity to biological output processes via

clock-controlled genes (Lee et al., 2001). Various expression profiling studies have

shown that, depending on the tissue, up to 10% of the transcriptome is under

circadian control (Akhtar et al., 2002, Bozek et al., 2009, Panda et al., 2002, Storch

et al., 2002). Accordingly, it is not surprising that long lasting circadian rhythm

disturbance (CRD), e.g. by chronically alternating light cycles, result in a loss of

homeostasis in addition to the acute disturbances as shown in several studies

(Archer et al., 2014a, Barclay et al., 2012, Moller-Levet et al., 2013).

As a consequence of chronic circadian rhythm disturbance (CRD), long-term shift

workers are at risk for various adverse health effects. Various studies indicated an

association between social jet-lag, shift work and increased body weight, although

conflicting data exist (Kubo et al., 2011, Nabe-Nielsen et al., 2011, Roenneberg et

al., 2012, Parsons et al., 2014). These human studies, however, are hampered by

difficulties with exposure assessment and confounding factors, such as unhealthy

food intake. Animal studies can limit the influence of these factors, as specifically

designed CRD exposure protocols allow exclusion of or strict control over

confounding factors. Previously, we have shown that chronically alternating light

cycles cause an increased breast cancer risk and body weight gain in breast cancer

predisposed mice (Van Dycke et al., 2015). This model provides a unique tool to

further analyze mechanisms that underlie the relationship between shift work and

cancer and increased body weight.

Shift work involves a multitude of aspects, including internal desynchronization,

melatonin suppression, sleep disruption, lifestyle disturbance and decreased

vitamin D levels (Fritschi et al., 2011). A number of these aspects, including circadian

desynchrony, lifestyle disturbance and timing of food intake, have been proposed

to underlie the relationship between shift work and increased body weight (Fonken


Chapter 6

100

et al., 2010, Salgado-Delgado et al., 2010). Unravelling the mechanisms that link

shift work with adverse health effects is a requisite for developing evidence-based

interventions that minimize health risks. Especially, understanding the underlying

metabolic changes could provide targets for preventive measures, since food

intake can be adjusted.

Transcriptomics studies provide a useful tool to investigate a wide variety of

processes and several studies show that CRD affects animal and human gene

expression profiles (Archer et al., 2014a, Barclay et al., 2012, Moller-Levet et al.,

2013). In human laboratory studies, mistimed sleep and insufficient sleep cause

the majority of the rhythmically expressed mRNAs in the blood transcriptome to

become arrhythmic (Archer et al., 2014a, Moller-Levet et al., 2013). Timed sleep

restriction in mice was shown to disrupt global liver gene expression, with many

genes associated with lipid and glucose metabolism processes (Barclay et al., 2012).

Furthermore, in these studies the exposure period was limited to weeks, whereas

shift workers are involved in irregular working hours for many years. Although

these human and animal studies show (semi)acute effects on both circadian and

non-circadian biological processes, directly or shortly after impingement on the

circadian system, they fail to provide insight into the constitutively disturbed

processes underlying long-term health effects such as obesity and cancer.

Therefore, to gain more knowledge on processes involved in chronic CRD we

performed a mouse study in which we chronically altered light cycles and focused

on changes in longer term expression patterns, which may explain the observed

weight gain phenotype.

Methods

Experimental set-upAnimal experiments were performed in compliance with national legislation,

including the 1997 Dutch Act on Animal Experimentation, and were approved by

the institute’s Animal Experimentation Ethical Committee. Food and water were

available ad libitum throughout the whole experiment.

The experimental set-up of the study has been described previously (Chapter 2)

(Van Dycke et al., 2015). In short, breast cancer-prone female p53R270H/+WAPCre mice



101

6

(Wijnhoven et al., 2005) in an FVB genetic background were kept under normal

LD conditions or chronically exposed to a weekly alternating 12:12 hour light-

dark cycle. After 18 inversions, on day 7 of the last shift, animals were sacrificed at

4 hour time intervals over a 24-hour period (n=4 animals per time point). Blood

and tissues were collected for further analysis. Due to premature animal loss, 21

animals per condition were available for further analysis. In an additional group of

mice (n=5 per group), a radio transmitter (Physio Tel, TA11 TA-F10; Data Sciences,

St. Paul, MN) was implanted in the peritoneal cavity to record locomotor activity

and core body temperature every ten minutes (Van Dycke et al., 2015).

Micro-array analysisTotal RNA was extracted from RNAlater (Invitrogen, Grand Island, NY, USA)

protected liver tissues using the miRNeasy Mini Kit (Qiagen Benelux, Venlo, The

Netherlands). RNA concentrations were measured using a NanoDrop ND-1000

Spectrophotometer (Nanodrop Technologies, Wilmington, DE, USA), and RNA

quality was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies,

Amstelveen, the Netherlands).

RNA was processed for gene expression analysis at the Microarray Department

of the University of Amsterdam, the Netherlands, using Agilent SurePrint G3

Mouse GE 8x60K arrays (Amstelveen, the Netherlands). This type of microarray

contains 55,681 probes for biological features such as mRNAs and ncRNAs. RNA

amplification and labeling were carried out according to the manufacturer’s

protocol. Experimental samples (each corresponding to RNA isolated from one

individual mouse) were labelled with Cy3 and the common reference sample

(made by pooling equimolar amounts of RNA from experimental samples) was

labelled with Cy5. Slides were scanned with an Agilent G2565CA DNA microarray

scanner, followed by automated feature extraction, resulting in a table containing

individual probe signal intensities for both dyes.

Data analysis

The raw data were subjected to a set of quality control checks to ensure comparable

signal average and distribution. Raw microarray data for gene-coding probes were

normalized in R (www.r-project.org) using a four step approach (Pennings et al.,

2011): (1) natural log-transformation, (2) quantile normalization of all scans, (3)

correcting the sample spot signal for the corresponding reference spot signal and


Chapter 6

102

(4) averaging data from replicate probe spots. The normalized data of 55,681 probes

were further analyzed in R and Excel (Microsoft Corporation, USA). Probe to gene

annotation data was downloaded from NCBI. Significantly regulated probes were

annotated to the corresponding genes for further biological interpretation. To this

end, when multiple probes corresponding to the same gene were significant, they

were counted as one gene in further analysis; probes that did not correspond to

genes according to the current NCBI database were excluded from further analysis.

To identify differentially expressed genes for chronic CRD exposed animals

compared to controls, we performed a one-way ANOVA with Qlucore Omics

Explorer (Qlucore AB, Lund, Sweden) in which p<0.001 was considered statistically

significant. CircWave Batch v5.0 software (Hut, R., www.euclock.org/results/

item/circ-wave-batch.html) was used to analyze circadian rhythmicity of gene

expression under regular LD conditions and after prolonged exposure to weekly

LD-inversions. Enrichment of selected gene lists was analyzed with MetaCore GO

Processes from GeneGo, Inc. (http://www.genego.com/). Gene expression data

were imported into Metacore using their NCBI Entrez GeneID as identifier; therefore

only genes annotated with a GeneID in the current NCBI database were used

for pathway enrichment analysis. To ensure biological relevance, only processes

including between 10 and 1000 genes were taken into account. GO Processes

were considered significantly enriched with a p<0.001. Gene Set Enrichment

Analysis allowed us to identify significant up- or downregulation of enriched gene

sets (P<0.05 and FDR<0.25) (Mootha et al., 2003, Subramanian et al., 2005).

Results

Attenuation of circadian rhythmicity in gene expressionThe alternating light cycle protocol to chronically disturb the circadian rhythm

(CRD protocol) used in this study has previously been shown to shorten tumor

latency times, increase body weight and affect sleep timing (Van Dycke et al.,

2015). Although the weight gain phenotype is already apparent at week 6,

group differences only became significant after week 24. Given the overweight

phenotype, we analyzed possible transcriptional changes in the liver. Since we were

interested in processes that are involved in the observed overweight phenotype

early on, we analyzed gene expression 18 weeks after start of the exposure, before

the differences in body weight gain were significant. Clock and clock-controlled



103

6

genes, as well as temperature rhythms, were re-entrained 7 days after the final

light-dark shift whereas corticosterone serum levels showed a bimodal rhythm

upon chronic CRD (Van Dycke et al., 2015). This re-entrainment of the majority

of these classical parameters allowed investigation of more permanent disturbed

genes and processes.

The observed re-entrainment of temperature and clock gene expression, in

contrast to altered serum corticosterone rhythms suggest that chronic CRD results

in internal desynchronization. Therefore, we determined circadian rhythmicity of

all genes using CircWave, for the control light-dark (LD group) and chronic CRD

groups separately. Significant circadian gene expression (p<0.001) was observed

for 594 genes in LD controls and only in 133 genes in chronic CRD exposed animals.

The overlap between the two groups was limited to 33 genes, including several

clock or clock controlled genes (for a complete overview, see Supplemental Table

S1). Interestingly, 561 genes lost significant circadian expression after chronic CRD

based on p-value. However, closer examination of the expression profiles revealed

that oscillation was not completely abolished for these genes (Fig. 1). Rather, the

majority of these genes showed a lower amplitude oscillations, as compared to

LD controls (Fig. 1) and the average amplitude of all 561 genes was significantly

decreased in chronic CRD exposed animals (Fig. 2a, Mann Whitney test, p < 0.001).

This lower amplitude explains the lack of significant circadian rhythmicity in gene

expression. Furthermore, a limited effect on phase was found (Fig. 2b). These

findings are indicative of a re-entrained, yet attenuated circadian rhythmicity in

gene expression upon prolonged exposure to weekly LD-inversions.


Chapter 6

104

Figure 1. Heatmap of 561 genes that lost significant circadian rhythmicity in gene expression after prolonged exposure to CRD, according to CircWave analysis. It is clearly shown that the circadian rhythm of these genes is not abolished, however the amplitude is decreased in the majority of these genes.



105

6

Figure 2. Overview of circadian parameters of genes with significant circadian gene expression under regular LD conditions, lost upon chronic CRD. A. The amplitude of these genes was significantly reduced in chronic CRD animals compared to controls B. Acrophase was comparable between no large phase shifts were found. All indicative of an attenuated circadian gene expression resulting from chronic exposure to CRD.

Mice exposed to weekly LD-inversions showed a larger increase in relative body

weight compared to the animals kept in a stable LD cycle (Van Dycke et al., 2015).

To gain insight in mechanisms underlying the observed increased weight gain,

we determined enrichment of the 561 genes that lost significant circadian gene

expression upon chronic CRD exposure compared to LD conditions in biological

processes. Chronic CRD caused attenuation of circadian rhythmicity in genes

involved in 237 processes (p <0.001, >10 and <1000 genes in process). The

largest groups of processes involve hormonal regulation, response to exogenous

compounds and metabolism, comprising both lipid and glucose metabolism (Fig.

3). In addition, the affected genes were associated with circadian rhythm, cell cycle,

and signaling processes (Fig. 3). A complete list of significantly enriched processes

is given in Supplemental Table S2.


Chapter 6

106

Figure 3. Significantly enriched processes for genes with attenuated circadian rhythmicity in gene expression upon chronic CRD exposure, arranged in themes. In relation to the observed overweight phenotype it is interesting to see decreased amplitude in genes involved in glucose and lipid metabolism processes, especially in glucose transport and storage and also lipid storage.

In addition to chronic CRD affected biological processes, we investigated the top

20 rhythmically expressed genes with the largest relative decrease in amplitude

(Supplemental Table S3) in more detail. Since we study gene expression changes

7 days after the last light-dark shift, only minor differences might be expected due

to re-entrainment. Within this group of most affected genes regarding amplitude,

several metabolic related genes still showed an altered rhythm, including Ppard,

Pdk4, Slc16a5 and Igfbp-1 (Fig. 4). Under LD conditions, both the nuclear hormone

receptor Ppard and isoenzyme Pdk4 peak at the dark to light transition. Chronic

CRD exposure appears to lower peak mRNA expression of these genes, and delay

acrophase to early light phase. Igfbp-1, which binds IGF and thereby regulates

IGF action, shows both alteration of phase and attenuation of the rhythm. Under

normal LD conditions, Igfbp-1 gene expression peaks at the dark to light transition,

whereas chronic CRD caused a slightly advanced and lower peak expression.

Moreover, the circadian rhythmicity of monocarboxylate transporter Slc16a5 gene

expression is completely abolished in chronic CRD exposed animals.

Disturbance of non-circadian genes and processesIn addition to disturbance of circadian rhythmicity of gene expression, chronic

disturbance of the circadian system might also affect non-circadian processes.

Using one-way ANOVA, we compared the overall expression of each gene of CRD

mice to that of the LD controls, not taking into account time-of-day. The latter was

accomplished by combining all time-points per group, comparing the average

expression of all CRD exposed mice to the average expression of all LD controls.

Surprisingly, we found only 18 genes to be differentially expressed (p <0.001) in

CRD exposed animals, as compared to controls (Supplemental Table S4), with



107

6

Figure 4. Altered circadian gene expression of Ppard, Pdk4, Igfbp-1 and Slc16a5, involved in the balance between glucose and lipid metabolism. Ppard and Pdk4 both show a phase delay and decreased amplitude upon chronic CRD (open symbols) compared to LD controls (closed symbols). CRD caused a slight advance and decrease of peak Igfbp-1 gene expression, with a distinct trough at the light to dark transition. For Slc16a5 the rhythm is completely abolished in CRD exposed animals.

minor diff erences in average gene expression. This number is limited, and does

not allow pathway analysis. Therefore, Gene Set Enrichment Analysis (GSEA) was

used, to identify subtle diff erences in GO processes or canonical pathways, again

using all time-points per group combined. In the GSEA analysis, the complete

gene expression dataset is used to study signifi cant overrepresentation of up

or down regulated genes between CRD exposed animals and LD controls in

gene sets representing GO Processes and other online databases (C2 gene sets).

GSEA identifi ed 36 upregulated processes or pathways, mainly involved in DNA

replication, transcription, translation and metabolism (Supplemental Table S5).

The majority of the 75 downregulated processes identifi ed represented immune,

cell cycle and proliferation processes (Supplemental Table S5). Interestingly,

also downregulation of several catabolic processes was observed. This analysis

indicates that although no major diff erences in average gene expression between

CRD animals and LD controls were found, genes involved in a wide variety of

processes are aff ected by CRD.


Chapter 6

108

Figure 5. Using Gene Set Enrichment Analysis significant overrepresentation of regulated genes between CRD exposed animals and LD controls in GO processes or the C2 gene sets is studied. The majority gene sets with significant overrepresentation of down regulated genes is involved in immune processes, whereas the up regulated genes are mainly involved in DNA replication.

Discussion

Due to the growing 24/7 economy, shift work will become increasingly part of our

society and will, therefore, have an incremental effect on public health outcomes.

Understanding mechanisms underlying the adverse health effects originating

from chronic circadian rhythm disturbance (CRD) is of prime importance to

develop preventive measures for shift work. An experimental set-up inducing

CRD, along with a mouse model at risk for the development of breast cancer

risk, weight gain and other morbidities, provides a unique tool to identify such

underlying disturbed processes (Kubo et al., 2011). The major advantage of this

model is the ability to control the exposure and to exclude confounding factors,

which both hamper human observational studies. Importantly, we aimed to find

constitutively disturbed processes upon long-term exposure to CRD. As such, we

studied the hepatic transcriptome of mice exposed to 18 weeks of CRD compared

to LD controls, 7 days after the last shift. We focused on identifying disturbances in

both circadian and non-circadian genes and processes.

Studying rhythmically expressed genes, we observed weaker rhythmicity in hepatic

gene expression upon chronic CRD exposure, affecting a wide variety of molecular

processes. This could be indicative for an overall loss of homeostasis, resulting from

constant re-entrainment to the chronically alternating light cycle. Although this



109

6

finding is generally in line with previous studies that have shown a loss of circadian

rhythmicity in the human blood transcriptome resulting from insufficient sleep

(Moller-Levet et al., 2013) and mistimed sleep (Archer et al., 2014a), our findings are

less pronounced. This can be explained by the fact that we monitored effects only

7 days of recovery after the last shift. This resulted in re-entrainment of classical

circadian parameters such as clock gene expression and core body temperature.

The loss of significant circadian gene expression is mainly due to a decrease in

amplitude, indicating that overall circadian gene expression is also re-entrained,

but rhythmicity is attenuated. In short-term experiments as performed by others

(Barclay et al., 2012), melatonin, clock genes and clock controlled genes were

found to be disturbed, directly after impingement on the circadian system.

In relation to the weight gain phenotype observed in mice under CRD conditions,

we found that genes with attenuated circadian expression upon CRD were

enriched for processes involved in mainly glucose but also lipid metabolism.

Under normal conditions both carbohydrate and lipid metabolism show distinct

time-of-day dependent rhythms (Bailey et al., 2014). The attenuation in glucose

metabolism processes is in line with recent findings that the circadian system

is largely responsible for daily rhythms in glucose tolerance and that separately

circadian misalignment reduces glucose tolerance (Morris et al., 2015). Furthermore,

GSEA analysis revealed down regulation of catabolic processes, which shows that

energy metabolism is also affected in non-circadian genes.

The largest decrease in amplitude was found in genes involved in the balance

between lipid and glucose metabolism: Ppard, Pdk4, Slc16a5 and Igfbp-1. The

nuclear receptor Ppard has been shown to be a key regulator of Pdk4 (Degenhardt

et al., 2007). This direct regulation could explain the similar changes in circadian

expression of both genes. A similar decrease in amplitude of Pdk4 expression was

observed in animals kept on a high fat diet (Eckel-Mahan et al., 2013), and animals

that underwent timed sleep restriction (Barclay et al., 2012), both indicative of an

interaction between circadian clock and metabolism. Furthermore, Pdk4 plays an

important role in the metabolic flexibility under various nutritional conditions

(Zhang et al., 2014b). In response to starvation Pdk4 mRNA and protein levels in liver

increase (Wu et al., 2000). The conversion of pyruvate into Acetyl-CoA is inhibited

by Pdk4 as it inactivates the pyruvate dehydrogenase complex, maintaining blood

glucose levels under fasting conditions (Jeoung and Harris, 2010). The observed


Chapter 6

110

circadian rhythm in Pdk4 expression under LD conditions correspond with daily

rhythms in feeding and fasting. Attenuated rhythms in Pdk4 resulting from chronic

CRD exposure might result in limited ability to regulate blood glucose levels.

Little is known about the function of Slc16a5, and mainly described as a

transporter for diuretics (Murakami et al., 2005). However, other monocarboxylate

transporters are involved in energy balance processes (Lengacher et al., 2013).

The Insulin-like Growth Factor-1 (IGF-1) axis, including Igpbp-1, is likely involved

in glucose metabolism, as it can acutely regulate glucose levels (Katz et al., 2002).

The combination of decreased amplitudes in Pdk4, Ppard, Igfbp-1 and Slc16a5

and increased body weight gain upon long-term exposure to CRD suggests an

imbalance between nutritional status and lipid and glucose metabolism, resulting

in obesity. Future studies should specify this imbalance and investigate which

interventions could help balance nutritional intake and metabolism to prevent

obesity. For example studying the metabolome could indicate which metabolic

processes have increased or decreased activity based on the metabolites found to

be linked to these processes.

In conclusion, chronic CRD results in weaker circadian rhythmicity of a wide variety

of circadian and non-circadian processes, including processes that could be related

to the body weight gain phenotype observed in shift workers. Attenuation in

circadian rhythmicity of both glucose and lipid metabolism processes and genes

involved in their balance provide a preliminary target for intervention. However,

the chronic effects are minor compared to the acute effects shown by previous

transcriptome analyses (Barclay et al., 2012, Moller-Levet et al., 2013). Taken

together, based on these observations one could speculate that the weakened

circadian rhythmicity 7 days after the last shift is only a remainder of more severe

disturbance at earlier time points e.g. 1 or 2 days after the shift. The relevance of

acute and more permanent disturbance of biological processes for adverse health

outcomes remains to be determined. Insights herein would provide important

input for the discussion whether slow or fast rotating shift schedules should be

employed. Additionally, in depth analysis of affected genes and processes and

metabolome data should provide further insights in the role of the constitutively

disturbed processes in the relation between CRD and increased weight gain.


CHAPTER 7

Summary and general discussion


Chapter 7

114

Summary

Working around the clock as a consequence of our growing 24/7 economy strains

and disturbs the tightly regulated circadian system. In epidemiological studies,

shift work was associated with many adverse health effects, such as cancer, obesity

and cardiovascular disease. These observational human studies are hampered by

confounding factors, difficulties with exposure assessment and take many years to

complete. Yet, due to the similarities in carcinogenesis (Balmain and Harris, 2000)

and the circadian system (Lowrey and Takahashi, 2011) between mice and human,

mice provide a unique tool to study the relationship between shift work and

cancer. Animal studies are less influenced by confounding factors and therefore

provide the ability to study the causal relationship between chronic circadian

disturbance and adverse health effects. Additionally, animal studies provide the

ability to study the different aspects of shift work and to identify underlying

mechanisms and potential biomarkers for circadian rhythm disturbance (CRD) to

aid the development and study of preventive measures.

In chapter 2 the effect of chronically alternating light cycles on breast cancer

development and body weight was studied using a unique breast cancer-prone

mouse model. We observed alterations in the classical circadian markers body

temperature and serum corticosterone, preceding health effects. Whereas age-

matched breast cancer prone Li-Fraumeni female mice, housed under normal light

dark conditions, displayed a tumor latency time of 50.3 weeks, animals exposed

to weekly light-dark (LD) inversions showed a markedly shortened tumor latency

time of 42.6 weeks. In addition, CRD animals showed an increase in body weight. To

our knowledge this is the first study providing experimental evidence for a causal

relationship between circadian rhythm disturbance and breast tumorigenesis.

These findings warrant the identification of biomarkers that evaluate the presence

of circadian rhythm disturbance (CRD) and health risk before adverse health effects

occur. The use of whole genome gene expression analysis is a valuable approach

to identify differentially expressed genes upon CRD in the tissue of interest and

select genes that qualify as useful biomarkers. In chapter 3, we provided a proof

of principle for a comparative transcriptomics approach to identify potentially

blood detectable biomarkers for breast cancer using the Li-Fraumeni mouse

model. By comparing gene expression profiles of tumor samples with different

characteristics to control mammary gland tissue, we determined differentially


115


expressed gene across tumor subtypes. Subsequently, candidates were prioritized

based on possible blood-detectability as determined by cross-referencing with

literature databases. Finally, a set of 16 candidate markers for breast cancer was

selected based on the expression in tumor compared to normal mammary gland.

As the transcriptomics approach proved to be able to select potentially blood-

detectable biomarkers for breast cancer, we employed a similar approach to identify

biomarkers for CRD. Although blood-detectable biomarkers for CRD, independent

of time of day, can facilitate studies regarding preventive interventions, these

biomarkers have not yet been identified. In chapter 4, we exposed female mice

to 6 shifts in a clockwise (CW) and counterclockwise (CCW) rotating CRD protocol.

Hepatic gene expression analysis using a classification approach resulted in a

consensus set of 15 genes, able to predict CRD with an accuracy ranging from 90%

to 98%. Additionally, to find candidate markers that allow non-invasive detection

of CRD, we applied a sequential approach selecting 9 potentially blood-detectable

markers. One of these candidates, CD36 was found to be significantly increased

in serum, although less pronounced than expected from the gene expression

upregulation.

Future research into the adverse health effects of CRD will also require the

measurement of disease related biomarker in large-scale cohort studies.

Chapter 5 provides a descriptive study of the daily variation of hormonal and

lipid biomarkers in males and females, in a molecular epidemiology-like setting.

Using two analysis methods: 1) Cosine fitting analysis and 2) RM-ANOVA, we found

diurnal variation in many markers for at least one gender, which exceeded the

inter-individual variation. Consequently, when investigating these markers one

should take into account diurnal variation. Furthermore, since time-of-day likely

matters for additional markers, these findings warrant the need for investigating

diurnal variation before setting up studies measuring markers in routine and

controlled settings.

In chapter 6 we describe experiments where mice were exposed to weekly

alternating light-dark cycles and the liver transcriptome was analyzed to identify

disturbed processes involved in the body weight gain phenotype as observed in

chapter 2. We found an overall attenuation of circadian rhythmicity in hepatic

gene expression, which was less pronounced than previous studies investigating


Chapter 7

116

the effect of insufficient or mistimed sleep on the blood transcriptome. Amplitude

was particularly decreased in genes involved in the balance between lipid and

glucose metabolism: Ppard, Pdk4, Slc16a5 and Igfbp-1. Thus, compromised

metabolic flexibility might underlie the body weight gain phenotype observed in

chapter 2 and the overweight phenotype in shift workers and therefore provide a

preliminary target for intervention.


117


Discussion

The existence of our 24-hour economy is irreversible and will require an increasing

number of individuals to work at irregular hours. This might put them at risk

for adverse health effects by constantly straining and disturbing the tightly

regulated biological clock. The relationship between the circadian system and

cancer has been studied extensively, both in epidemiological studies and animal

experiments. Most epidemiological studies point towards an association between

shift work and breast cancer, as well as other adverse health effects (IARC, 2010).

Unfortunately, the complex mixture of exposures we call shift work hampers the

accurate assessment of causality. Furthermore, large scale prospective studies are

costly and labor intensive due to the use of classical circadian markers such as

melatonin, requiring around the clock measurement. To date, animal studies have

emphasized the relationship between the circadian system and cancer. Overall,

they have shown that a partially or completely non-functional circadian system

result in increased tumor growth. However, studies are lacking that combine

relevant models for CRD exposure and adverse health effects such as breast cancer

to provide experimental proof for the shift work cancer connection. Notably, there

is a gap between epidemiological and animal studies concerning the experimental

proof that shift work causes an increased cancer risk. In this thesis we aimed to

elucidate the causal relationship between circadian rhythm disturbance and

breast cancer development and to provide candidate biomarkers and targets for

interventions for future studies.

Causality between CRD and breast cancer and obesity The findings described in chapter 2 show that chronic circadian rhythm

disturbance (CRD) induced by chronically alternating light schedules decreased

latency times of tumor development and increased body weight gain in a

breast cancer-prone mouse model. Previous animal experiments, studying the

relationship between CRD and cancer differed from this study in the combination

of CRD induction methods and the breast cancer models used. For example, some

studies addressed tumor growth rather than cancer initiation and progression

by using xenograft models (Blask et al., 2003, Filipski et al., 2004) and chemically-

induced tumor models (Hamilton, 1969, Shah et al., 1984). Additionally, ‘unnatural’

continuous bright light exposure and endogenous circadian rhythm disruption

models were employed, which result in arrhythmicity (Anisimov et al., 2004, Wu


Chapter 7

118

et al., 2011). To our knowledge, the study as described in chapter 2 unequivocally

shows a link between chronic CRD and breast cancer development, relevant for

the human situation.

Shift work is a complex mixture of exposure and different aspects of these exposures

have been suggested to underlie the observed cancer risk in individuals involved

in shift work. Fritschi et al. proposed five mechanisms that should be considered

for comprehensive investigation: (i) internal desynchronization, (ii) light at night

(resulting in melatonin suppression), (iii) sleep disruption, (iv) lifestyle disturbances

and (v) decreased vitamin D levels due to lack of sun light (Fritschi et al., 2011). In

chapter 2 the relevance of these hypotheses was tested and the findings suggest

that internal desynchronization and sleep disruption are important in the etiology

of CRD-associated health risks and pathologies. The other factors (melatonin

suppression, lowered vitamin D levels and life style factors) could be excluded

in our study and therefore appear to be less relevant in the shift work cancer

relationship. Consequently, sleep disruption and internal desynchronization are

targets for preventive measures.

Apparently, the process of carcinogenesis is induced by circadian rhythm

disturbance and additional experiments could further elucidate which specific

aspects of carcinogenesis is affected. It has been suggested that all aspects of the

cellular response to DNA damage are controlled or influenced by the circadian

clock (Sancar et al., 2010). Consequently, chronic disturbance of the circadian

system is likely to affect these DNA damage control systems in place to prevent the

accumulation of mutations, which eventually result in uncontrolled or neoplastic

cell growth (Berenblum, 1975). Alternatively, the accumulation of mutations could

be the result of increased exposure to endogenous DNA damaging agents. For

example, the disturbance of metabolic processes by CRD could increase the

amount of reactive oxygen species and consequently increase mutations. Based

on the decreased latency time to tumor development observed in chapter 2

one could hypothesize that the number of mutations increased faster under

CRD conditions compared to LD conditions. This hypothesis can be addressed by

evaluating the mutation frequency in vivo using a plasmid based transgenic mouse

model (Dolle et al., 1996). Besides DNA repair processes, the immune system also

plays an important role in breast tumorigenesis (Standish et al., 2008), and circadian

disturbance has been shown to affect the immune response (Phillips et al., 2015),


119


increasing vulnerability for environmental stressors. Furthermore, epigenetic

changes resulting from prolonged exposure to CRD could lead to increased tumor

development, for example by silencing of tumor suppressor genes.

Besides the increased risk of cancer, epidemiological studies have also indicated

an association between social jet-lag, shift work and body weight gain, although

again conflicting data exist (Kubo et al., 2011, Nabe-Nielsen et al., 2011, Roenneberg

et al., 2012, Parsons et al., 2014). Various mouse studies and the findings described

in chapter 2 have shown that circadian disturbance induced by continuous

light, forced activity or a disrupted LD cycle causes an increase in body weight

(Coomans et al., 2013, Oishi, 2009, Salgado-Delgado et al., 2008). In the liver, a key

organ in glucose and lipid metabolism, approximately 10% of the transcriptome is

rhythmically expressed, including metabolic processes (Akhtar et al., 2002, Panda

et al., 2002). Furthermore, under normal conditions both carbohydrate and lipid

metabolism show distinct time-of-day dependent rhythms (Bailey et al., 2014).

We hypothesized that long-term CRD results in more permanent alterations in

biological processes relevant for the observed phenotype.

In chapter 6, using the transcriptomics technique, we found an overall attenuation

of circadian rhythmicity in gene expression 7 days after the last shift. The overall

attenuation of circadian gene expression is similar, although less pronounced,

in comparison with other studies using an omics approach to find the effects of

disrupted sleep (Archer et al., 2014a, Barclay et al., 2012, Moller-Levet et al., 2013).

The main difference between our study and others is the timing when the effects

are observed. Where others studied circadian rhythmicity of gene expression

directly after impingement on the circadian system, we studied the transcriptome

7 days after the last shift of long-term CRD exposure. One could speculate that the

weakened circadian rhythmicity 7 days after the last shift is only a remainder of

more severe disturbance at earlier time points e.g. 1 or 2 days after the shift. Follow-

up studies should further investigate whether the number of genes that show an

attenuation in circadian rhythmicity in gene expression and the magnitude of the

amplitude decrease correlate with the magnitude of CRD and subsequent health

effects.

The largest decrease in amplitude was observed in genes involved in the balance

between glucose and lipid metabolism. Further research to find the exact


Chapter 7

120

imbalance in metabolism could provide an interesting target for intervention,

with timed nutritional intake. Previous animal studies have shown that timed

feeding can indeed (partially) rescue tumor and obese phenotypes (Filipski et al.,

2005, Fonken et al., 2010, Oike et al., 2015). In all cases food was restricted to the

active phase and was shown to prevent dim light or jet lag induced body weight

gain and accelerated xenograft tumor growth by chronic jet lag. The finding that

advances and delays in meal-timing affect both circadian rhythms as well as

metabolic parameters underlines the importance and potential of meal-timing

(Yoon et al., 2012).

Evidence based preventive measuresDue to the high incidence of breast cancer and obesity an increased risk resulting

from long-term shift work can have major public health implications. However, shift

work is irreversibly linked to our 24/7 society and consequently evidence based

interventions to prevent CRD and consequent health effects are needed. Currently

advised preventive measures are versatile, including controlled light exposure,

schedule adaptations, and behavioral and pharmaceutical interventions (Neil et

al., 2014) and underlying philosophies and results are contradictive. Interventions

using controlled light exposure are aimed at phase shifting the circadian rhythm

and promoting adaptation to work at night. For example, by exposure to bright

light during the night shift and wearing dark goggles on the commute home in

the morning. This was shown to effectively increase day-time sleep (Boivin et al.,

2012, Sasseville and Hebert, 2010) but also without effects on sleep (Budnick et

al., 1995). In contrast, changing schedules from slow backwards rotating to fast

forward rotation, are based on the principle of limited circadian rhythm adjustment

(Knauth, 1996). Implementation of this fast forward rotation schedule has shown

contradicting effects on sleep and metabolic markers (Boggild and Jeppesen,

2001, Harma et al., 2006, Viitasalo et al., 2008). A major limitation of these studies

is the relatively short follow-up and lack of association with adverse health effects.

It is important to acknowledge that the positive effects of interventions on the

acute shift work related health complaints, such as sleepiness and gastrointestinal

symptoms, do not necessarily represent preventive effects on long-term health

effects, such as cancer and obesity.

The importance of preventive measures is also acknowledged by the Dutch

government, which requested the Dutch Health council to advice on the availability


121


of preventive measures (Health Council of the Netherlands, 2015). Based on the

afore mentioned review (Neil et al., 2014) and two additional reviews (Ruggiero

and Redeker, 2014, Liira et al., 2014), it was concluded that there are no evidence

based preventive measures to be recommended for shift workers to suppress

long-term health effects. The council expresses the need for both experimental

and observational studies that link suggested preventive measures with long-term

health effects.

The CRD model in combination with the breast cancer-prone mouse model as

used in chapter 2, are an ideal starting point to identify evidence based preventive

measures that decrease long-term health risk. Firstly, because of the relative short

period needed to evaluate the effect of these interventions on long-term health

effects compared to human studies. And secondly, the ability to continuously

measure activity and core body temperature using a telemetry system, enables

the monitoring of circadian system adaptation. Cross-sectional studies or use of

reporter mice for clock genes can evaluate whether peripheral clock also adapt,

similar to locomotor activity and core body temperature rhythms. (Chaves et al.,

2011). Hereby, providing experimental evidence whether interventions promoting

adaptation or limiting adjustment of the circadian system are best suited to

prevent disease. Furthermore, sleep patterns can be determined using mouse

telemetry data as described previously for human wrist actimetry data (Juda et al.,

2013b). This is important since sleep duration, timing and quality are often used

to describe shift work-related strain (Akerstedt, 2003) or used as post-intervention

measures (Neil et al., 2014, Vetter et al., 2015). Overall, specialized animals can be

used to screen the variety of preventive measures proposed (Neil et al., 2014) and

experimentally proven interventions can be studied in shift workers.

BiomarkersAs described above, many recommendations have been made to alleviate shift

work related health effects. Although animal models can be used to identify

interventions that are likely preventing adverse health effects, there is a need for

scientific evaluation of these preventive measures in shift workers (Richter et al.,

2010). Tools that measure the presence of CRD or evaluate the health risk could aid

the evaluation of new interventions and provide individual, longitudinal exposure

assessment in shift workers. In the context of shift work, two types of biomarkers

can be differentiated: 1) biomarkers that report on the presence and magnitude of


Chapter 7

122

circadian disturbance and 2) biomarkers that are associated with disease. The latter

do not report on CRD, but can identify health risk which allows early intervention

before adverse health effects occur.

Mouse models are frequently applied in biomarker discovery research (Kelly-

Spratt et al., 2008). Similarities exist between breast cancer mouse models and

human breast cancer, and these mouse models are therefore useful in biomarker

discovery studies (Klein et al., 2007, Pitteri et al., 2008). In chapter 3 the Li-Fraumeni

mouse model mimicking human breast cancer development was used to provide

proof of principle for the use of transcriptomics in the identification of new blood-

based biomarkers for breast cancer. In this study, we used a genomics approach

on mammary gland tumor tissue compared to find genes overexpressed in tumor

tissue compared to control. Blood detectability was determined based on Gene

Ontology ‘extracellular’ or UniProt ‘secreted’ annotation, or cross-reference with the

Human Plasma Proteome project (Anderson et al., 2004). Additional selection based

on actual expression in human breast tumor compared to normal mammary gland

tissue yielded a set of 16 promising biomarkers for further validation. Importantly,

circadian rhythmicity and food intake responsiveness should be excluded.

After validation, a panel of blood-based biomarkers for breast cancer as identified

by this approach appeared a valuable add-on to currently available early detection

methods for breast cancer screening. Using sensitive detection techniques, blood-

based biomarkers can be detected in small aliquots of blood obtained by a finger

prick (Pennings et al., 2014), which certainly is less invasive than undergoing

mammography. Furthermore, blood-based biomarkers are not hampered by high

density breast tissue, as is the case with mammography. This way, measurement

of biomarkers is minimally invasive with relatively low costs, which makes them

applicable for use in a large scale population screening programs. People at high

risk of breast cancer, including shift workers, could be eligible for a blood test

using such a panel of early detection biomarkers. Besides prevention, the most

efficient way to reduce cancer mortality and morbidity is detection at an early

stage, allowing effective therapeutic treatment. In the future, the breast cancer

biomarker panel can even be combined with early detection markers for other

tumor types in a multiplex assay.


123


Upon the proof of principle for a transcriptomics approach in the identification of

blood-detectable biomarkers for breast cancer, as provided in chapter 3, a similar

approach was taken to identify biomarkers for circadian rhythm disturbance in

chapter 4. In the future, these markers are intended for the use in large cohort

studies, such as the Nightingale Study (Pijpe et al., 2014) or the Doetinchem

Cohort Study (Verschuren et al., 2008). To be eligible for use in such settings,

these universal biomarkers should fulfill the following criteria: 1) independent

of rotation schedule; 2) independent of time of day; 3) age-independent and

4) blood-detectable. Recently, transcriptomics- and metabolomics-based time

of day independent biomarkers have been identified using jet lag exposed and

clock mutant mice (Minami et al., 2009, Ueda et al., 2004), allowing detection

of acute and transient effects of CRD, but not chronic effects. In chapter 4 we

identified an optimal CRD classification set based on hepatic gene expression and

a selection of potentially blood detectable biomarkers, both focused on time-of-

day independent detection of chronic CRD.

Although these selected blood-based biomarkers fulfill the criteria for use in large

scale cohort studies, they should be further validated. Importantly, researchers

currently using classical circadian markers need to be convinced of the relevance

of these markers for CRD and their usefulness in relation to adverse health effects.

Longitudinal measurements in both animal and human studies are required

to evaluate the association of these biomarkers with CRD exposure duration.

In addition to the relationship with exposure, the identified markers should be

predictive for CRD associated health risk. Previous studies have shown that

comparable chronic CRD protocols increased negative health effects in mice

(Davidson et al., 2006, Filipski et al., 2004). Validation of the biomarkers in human

studies are required to show whether this association with disease also applies to

humans.

As described above, one could also opt for the measurement of disease associated

biomarkers in a molecular epidemiology-like setting. Especially the combination

of biomarkers for CRD and biomarkers for the adverse health effects could identify

individuals at risk for disease development. Hormonal imbalance is often studied

in large epidemiological settings since these are associated with high incidence

cancer (such as breast cancer), and it is well known that metabolic markers might

serve as important (early) indicators for metabolic disorders (Mester et al., 2010,


Chapter 7

124

Stevens, 2005, Ferlay et al., 2015). In relation to shift work, metabolic markers such

as cholesterol and triglycerides have been used to evaluate the advantageous

effect of altered shift schedules (Orth-Gomer, 1982, Viitasalo et al., 2008). It is

known that several disease related biomarkers show a diurnal rhythm in serum

or plasma levels. However, most information on diurnal variation is derived from

these controlled laboratory studies in which factors influencing diurnal variation

(such as food intake and/or sleep/wake cycle) are controlled (for example see

references (Klingman et al., 2011, Li et al., 2013, Jung et al., 2010, James et al., 2007,

Davies et al., 2014, Ang et al., 2012, Dallmann et al., 2012)). With limited control

over environmental factors in large scale cohort studies, it is important to have

insight into the daily variation of these markers in a routine non-fasted molecular

epidemiology-like setting, with no restrictions on sleep or diet.

In chapter 5 we have shown that the majority of the hormonal and metabolic

markers studied showed a daily variation (circadian rhythms and/or time-of-day-

effects that does not fit a cosine curve) in at least one gender. A recent study also

showed that 43% of the protein coding genes show diurnal variation in at least

one organ in mice (Zhang et al., 2014a), this indicates that daily variation in serum

or plasma proteins is likely. In large molecular epidemiological studies, it should

be advised to study diurnal variation of markers of interest and take into account

when present. Preferably, samples should be taken at one time point. Furthermore,

time of sampling and relevant environmental factors, such as food intake, can be

registered.

In shift workers, the measurement of markers with circadian rhythmicity is further

complicated by variation caused by circadian disturbance, altered sleep timing

and food intake and other shift work related factors. Consequently, when variation

within groups is increased due to random sampling or disturbance by work

schedule, larger group sized are required to increase power do detect differences

between groups. To limit the additional variation, samples of biomarkers with

diurnal variation in shift workers should best be taken on the last day shift. This

way, individuals are least disturbed by work schedule and mostly recovered from

previous shifts. Although this depends on chronotype, impingement on sleep

wake rhythms by work hours and social responsibilities is limited. And it take

several days for disturbed circadian rhythms to reentrain. Preferably, finger prick

blood samples, urine samples and saliva samples, when applicable for the marker


125


of interest, could be taken at home on free days without the interference of work

schedule.

Modelling shift work in animal studiesIn this thesis, light schedules were altered to cause circadian rhythm disturbance.

We have shown that altered light schedules in a mouse model for human breast

cancer can provide important insights about health effects observed in human

studies. Another important aspect of shiftwork is the activity at times of the day

when we would normally be asleep. In humans, this consequently leads to altered

meal timing and sleep disruption (Grundy et al., 2009, Lennernas et al., 1995). This

can be mimicked by gently handling the animals and exposing them to new

objects in their environment during normal sleeping hours, referred to as Timed

Sleep Restriction (TSR) (Barclay et al., 2012). However, this labor intensive method

is not compatible with long-term studies in large cohorts of animals. Currently,

experiments are ongoing with specially developed ‘shift work cages’ in which

the animals are gently, but continuously forced to be active at predefined time

windows of the day (mimicking day, evening and night shifts). This set-up has been

validated with similar sleep analyses as described in chapter 2. In combination

with the Li-Fraumeni mouse model, the effect of altered sleep-wake cycle resulting

from rotating shift work on breast cancer risk can be addressed.

Although the increased breast cancer risk is probably the most extensively studied

cancer type associated with shift work, other types of cancer, such as colorectal

and prostate cancer have been prospectively been associated with long-term shift

work (Kubo et al., 2006, Schernhammer et al., 2003). Accurate exposure assessment

and control for confounding factors is also limited for these types of cancer. Using

mouse models for colorectal cancer (Karim and Huso, 2013) or prostate cancer

(Ittmann et al., 2013), the effect of chronic CRD, either induced by chronically

alternating light cycles or forced activity, on these cancer types can be studied. In

addition to the cancer phenotype, other disease pathologies such as diabetes type

2 and obesity can be studies in animal models. Similarly to cancer, epidemiological

research for these diseases is confounded by unhealthy habits of shift workers,

which can be controlled for in experimental studies.


Chapter 7

126

Future perspectivesThe fact that shift work has become an irreversible part of our 24/7 society

underlines the importance of unravelling which aspects of shift work are relevant

for the observed health effects and how to develop interventions to prevent

adverse effects. To further study the relationship between shift work and adverse

health effects, it is of utmost importance to inventory what shift work entails and

to study the relevance of these individual aspects. This requires the collaboration

of researchers with expertise of the circadian system and animal models and

researchers with the ability to provide detailed information on human shift work

characteristics, to close the gap between animal studies and epidemiological shift

work studies in humans. Detailed knowledge on light exposure, food intake, sleep

timing, duration and quality and other life style or biological characteristics from

shift workers can be used as input for laboratory animal studies. Cross-sectional

studies in shift workers and day workers can identify the aspects that are specific

for individuals working in shifts. Animal experiments allow the comparison of

these individual aspects of shift work or combinations in relation to health effects

in a relatively short time frame.

Alternatively, the findings from these animal studies can provide valuable input

for the set-up of prospective human research. Upon validation, the candidate

biomarkers identified in chapter 4 are interesting alternatives for the classical

markers previously used in human studies. Based on the findings in chapter 6

and those of others (Filipski et al., 2005, Fonken et al., 2010, Oike et al., 2015)

we hypothesized that the timing of food intake is important for both increased

body weight and cancer risk. Therefore, prospective human studies would ideally

register meal-timing in shift workers to associate this with obesity and cancer risk.

Additionally, animal studies are in progress to identify the ideal timing of food

intake in a simulated shift work setting in relation to long term health outcome.

The most promising findings of such experiments can provide valuable input for

food intake interventions to be validated and evaluated in cohort studies with shift

workers.


CHAPTER 8

Nederlandse Samenvatting Curriculum Vitae

List of publications PhD Portfolio

Dankwoord


130

Chapter 8

Nederlandse Samenvatting

In vrijwel alle organismen is een biologische klok aanwezig, van fruitvliegjes tot

mensen. De biologische klok wordt gedreven door de master klok in de hersenen,

de supra chiasmatische nucleus. Deze master klok houdt een ritme aan van

ongeveer 24 uur, ook wel circadiaan ritme genoemd, en wordt op tijd gezet door

licht. Lichaamstemperatuur en activiteit zijn voorbeelden van processen die door

deze master klok geregisseerd worden en dus over de dag variëren. Daarnaast zijn

er ook hormonen zoals melatonine en cortisocosteron die een ritme hebben van

ongeveer 24 uur. Moleculair gezien bestaat de klok uit verschillende genen die via

feedback mechanismen over de dag verschillend tot expressie komen.

Het werken in ploegendienst, als consequentie van onze groeiende 24/7 economie,

belast en verstoort het strak gereguleerde biologisch ritme. In epidemiologische

studies is het werken in ploegendienst geassocieerd met vele gezondheidsrisico’s,

zoals kanker, obesitas en cardiovasculaire aandoeningen. Deze observationele

humane studies worden gehinderd door verstorende factoren, moeilijkheden

met het vaststellen van de blootstelling en nemen daarnaast vele jaren in beslag.

Door de sterke overeenkomsten in het carcinogenese en circadiane proces tussen

muis en mens bieden muizen een unieke mogelijkheid om de relatie tussen het

werken in ploegendienst en kanker beter te onderzoeken. Dierstudies worden

niet beïnvloed door verstorende factoren en bieden daardoor de mogelijkheid

om de causale relatie tussen chronische circadiane verstoring en lange termijn

negatieve gezondheidseffecten te bestuderen. Bovendien kunnen de afzonderlijke

aspecten van ploegendienst arbeid onderzocht worden. Door de toepassing van

moleculair biologische technieken kunnen ook de onderliggende mechanismen

en potentiele biomarkers geïdentificeerd worden, welke worden gebruikt voor de

ontwikkeling en studie van preventieve maatregelen.

In hoofdstuk 2 wordt beschreven wat het effect is van chronische jetlag op

borsttumorontwikkeling en lichaamsgewicht, gebruikmakend van een unieke

muizen met aanleg voor borstkanker. Dit muismodel heeft een mutatie in de

borstklier die ook bij vrouwen met borstkanker voorkomt, hierdoor ontwikkelen

deze muizen borsttumoren die vergelijkbaar zijn met humane borsttumoren. Onder

normale licht-donkercondities ontwikkelden deze vrouwelijke, kankergevoelige

muizen een borsttumor na gemiddeld 50.3 weken. Wanneer dit type muis werd


131

8

Nederlandse samenvatting

blootgesteld aan een wekelijkse omkering van het licht-donkerritme ontwikkelden

borsttumoren significant eerder, al na gemiddeld 42.6 weken. Bovendien lieten ze

een toename in lichaamsgewicht zien. Voor zover bekend is dit de eerste studie

die experimenteel bewijs levert voor de causale relatie tussen verstoring van het

circadiane ritme en borsttumor ontwikkeling.

Deze bevindingen maken het noodzakelijk om biomarkers te identificeren

die de aanwezigheid van circadiane verstoring evalueren voordat negatieve

gezondheidseffecten zich openbaren. Hiervoor is het waardevol om de

expressie van alle genen te vergelijken tussen een controlegroep en een groep

die is blootgesteld aan circadiane verstoring (CRD). De genen die verschillend

tot expressie komen, kunnen als basis dienen voor de selectie van bruikbare

biomarkers. Het tegelijkertijd in kaart brengen van de expressie van alle genen

noemen we transcriptoomanalyse. In hoofdstuk 3 hebben we een proof-

of-principle gegeven voor het gebruik van transcriptoomanalyse om bij de

muis biomarkers te identificeren die potentieel in bloed te detecteren zijn.

Door de genexpressieprofielen van borsttumoren te vergelijken met controle

borstklierweefsel vonden we genen die differentieel tot expressie kwamen in

borsttumoren. Vervolgens werden kandidaat biomarkers gerangschikt op basis

van potentiële detecteerbaarheid in bloed door vergelijking met literatuur

databases. Uiteindelijk werden de 16 meest humaan relevante kandidaat markers

geselecteerd.

Tot op heden waren er geen in bloed detecteerbare biomarkers voor CRD

beschikbaar die onafhankelijk van het tijdstip van de dag gemeten kunnen

worden. Deze biomarkers zijn nodig om studies naar preventieve maatregelen

te faciliteren. Omdat transcriptoomanalyse bruikbaar bleek om biomarkers te

identificeren die potentieel in bloed detecteerbaar zijn, werd een vergelijkbare

aanpak gebruikt voor het vinden van biomarkers voor CRD. In hoofdstuk 4

onderwierpen we vrouwelijke muizen aan 6 lichtverschuivingen met de klok mee

(CW) of tegen de klok in (CCW). Genexpressie analyse van de lever resulteerde

in een set van 15 genen, waarmee we in staat waren om de aanwezigheid van

circadiane verstoring te voorspellen met een nauwkeurigheid van 90% tot 98%.

Daarnaast hebben we een stapsgewijze aanpak gebruikt om kandidaat markers te

vinden voor niet-invasieve detectie van CRD. Met deze aanpak selecteerden we 9

potentiële, in bloed detecteerbare markers. Eén van deze markers, CD36 was ook

in muizen serum significant verhoogd na CRD.


132

Chapter 8

Toekomstig onderzoek naar de negatieve gezondheidseffecten van CRD vereist

ook het meten van ziektegerelateerde biomarkers in grote cohort studies. Het is

voor een aantal van deze markers bekend dat ze een circadiaan ritme vertonen.

Echter is dit vaak onderzocht onder streng gecontroleerde condities, die niet

representatief zijn voor cohort studies. Hoofdstuk 5 geeft een beschrijving van

de dagvariatie van humane hormonale en lipide biomarkers in een setting met

beperkte controle over de condities die de variaties beïnvloeden. We vonden een

dagvariatie in veel markers die groter was dan de interindividuele variatie. Daarom

zou men bij het bestuderen van deze markers de dagvariatie in ogenschouw

moeten nemen. Bovendien is het aannemelijk dat het moment van de dag ook van

belang is voor additionele markers. Dit vereist dat voor het opzetten van nieuwe

studies met een beperkte controle over de meetcondities, vooraf de dagvariatie

van de te meten biomarkers bepaald wordt.

In de studie beschreven in hoofdstuk 6 zijn muizen blootgesteld aan wekelijks

alternerende licht donker cycli. Van deze dieren is het transcriptoom van de lever

geanalyseerd om verstoorde processen te identificeren die ten grondslag liggen

aan het toegenomen lichaamsgewicht, zoals geobserveerd in de dierstudie

beschreven in hoofdstuk 2. Een algehele afname van de circadiane ritmiek in

lever genexpressie werd gevonden, die minder uitgesproken was dan gevonden

in eerdere studies. De voornaamste afname in amplitude werd gezien in genen

betrokken bij de balans tussen vet en glucose metabolisme: Ppard, Pdk4, Slc16a5

and Igfbp-1. Het is dus mogelijk dat verstoorde metabole flexibiliteit ten grondslag

ligt aan het lichaamsgewichttoename fenotype gezien in hoofdstuk 2 en het

overgewicht in ploegendienst medewerkers en daarom dus een preliminair

aanknopingspunt voor interventie is.

De bevindingen beschreven in dit proefschrift tonen aan dat ploegendienstarbeid

negatieve invloed heeft op de gezondheid. Omdat ploegendienstarbeid door de

24/7 economie onlosmakelijk verbonden is aan onze maatschappij, vereist dit

preventieve maatregelen om de gezondheid van ploegendienstmedewerkers te

waarborgen. Op dit moment zijn er geen wetenschappelijk bewezen interventies

beschikbaar, mede doordat langdurig onderzoek nodig is om de effectiviteit

op lange termijn gezondheidseffecten te besturen. De ontwikkeling van deze

interventies kan gefaciliteerd worden door biomarkers die in een vroegtijdig

stadium ziekte of circadiane verstoring kunnen meten. Met behulp van


133

8

Nederlandse samenvatting

transcriptoomanalyse is in dit proefschrift de basis gelegd voor de identificatie van

deze biomarkers die onafhankelijk van het tijdstip van de dag gemeten kunnen

worden. Bovendien kunnen diermodellen op korte termijn waardevolle informatie

opleveren over de werkzaamheid van interventies bij mensen. Daarnaast hebben

we laten zien dat het voor ziektegerelateerde biomarkers belangrijk is om rekening

te houden met de dagvariatie bij het opzetten van humane studies. Ten slotte

hebben we metabole flexibiliteit geïdentificeerd als eerste aanknopingspunt voor

de benodigde interventies.


134

Chapter 8

Curriculum Vitae

Personal informationName Kirsten Catharina Gabriëlla Van Dycke

Date of birth 14 November 1986

Place of birth Arnhem

Education2007 – 2010 MSc. Biomedical Sciences, Radboud University Nijmegen

2004 – 2007 BSc. Biomedical Sciences, Radboud University Nijmegen

1998 – 2004 VWO, Nature & Health, Roelof van Echten College Hoogeveen

Research Experience2010-2015 PhD project

Department of Genetics, Erasmus MC, in collaboration with

the National Institute for Public Health and the Environment,

Prof. G.T.J. van der Horst and Prof. H. van Steeg

Working around the clock: Adverse effects of circadian

rhythm disturbance

2010 Thesis Consultancy profile

Dutch Health Council, Dr. H. van Dijk

The role of consensus in policy advising by the Health

Council of the Netherlands

2009 Master thesis Health Technology Assessment

McGill University Health Center, Canada, Prof. J. Brophy

Cost-effectiveness of high-sensitivity C-reactive protein

screening followed by targeted statin therapy in

asymptomatic individuals

2008 Master thesis Toxicology

National Institute for Public Health and the Environment, Dr. M.

Luijten

Effects of exposure route on gene expression profiles in the

mouse


135

8

Curriculum Vitae

2007 Bachelor thesis Health Technology Assessment

Comprehensive Cancer Centre East (IKO), Dr. M. de Boer

Cost-effectiveness of adjuvant systemic therapy in patients

with early stage breast cancer and micrometastases in the

sentinel lymph node


136

Chapter 8

List of publications

van Kerkhof LW*, Van Dycke KC*, Jansen EH, Beekhof PK, Ruskovska T, Velickova

N, Kamcev N, Pennings JL, van Steeg H, Rodenburg W. (2015) Diurnal variation of

hormonal and lipid biomarkers in humans. PLoS One 18;10(8)


Van Dycke KC, Rodenburg W, van Oostrom CT, van Kerkhof LW, Pennings JL,

Roenneberg T, van Steeg H, van der Horst GT. (2015) Chronically alternating light

cycles increase breast cancer risk in mice. Curr Biol 25(14):1932-7

Van Dycke KC, Pennings JL, van Oostrom CT, van Kerkhof LW, van Steeg H, van

der Horst GT,

Rodenburg W. (2015) Biomarkers for circadian rhythm disruption independent of

time of day. PLoS One 18;10(5)

Van Dycke KC*, Nijman RM*, Wackers PF, Jonker MJ, Rodenburg W, van Oostrom

CT, Salvatori DC, Breit TM, van Steeg H, Luijten M, van der Horst GT. (2014) A day and

night difference in the response of the hepatic transcriptome to cyclophosphamide

treatment. ArchToxicol 89(2):221-31


Pennings JL*, Van Dycke KC*, van Oostrom CT, Kuiper RV, Rodenburg W, de Vries

A. (2012) Biomarker discovery using a comparative omics approach in a mouse

model developing heterogeneous mammary cancer subtypes. Proteomics

13:2149-57


de Boer M, Adang EM, Van Dycke KC, van Dijck JA, Borm GF, Seferina SC, van

Deurzen CH, van Diest PJ, Bult P, Donders AR, Tjan-Heijnen VC. (2012) Cost-

effectiveness of adjuvant systemic therapy in low-risk breast cancer patients with

nodal isolated tumor cells or micrometastases. AnnOncol 10: 2585-91


137

8

PhD portfolio

PhD portfolio

Name PhD student: Kirsten Van DyckeErasmus MC Department: GeneticsResearch School: MGC graduate school

PhD period: 2010-2014Promotor(s): G.T.J. van der Horst & H. van SteegSupervisor: W. Rodenburg

1. PhD trainingYear Workload

(Hours)

General courses - Laboratory animal science- Literature discussion- Writing in English for Publication

201120112013

1202060

Specific courses (e.g. Research school, Medical Training)- CBS Chronobiology Summerschool 2012- Begin R-cursus- Transgenesis, Gene Targeting and in vivo

Imaging- Analysis of Microarray Gene Expression data

201220112011

2011

404

40

40


138

Chapter 8

1. PhD trainingYear Workload

(Hours)

Seminars and workshops- GZB meetings + work discussions- Center for Timing Research meeting 2011- Leopoldina Symposium (poster

presentation)- Center for Timing Research meeting 2012

spring (1-minute presentation)- Center for Timing Research meeting 2012

autumn (oral presentation)- MGC PhD student Workshop Düsseldorf

(poster presentation)- MGC PhD student Workshop Luxembourg

(oral presentation)- PhD days Nederlandse Vereniging voor

Toxicologie (NVT) (poster presentation)- MGC Symposium Rotterdam (oral

presentation)- Center for Timing Research meeting 2014

(oral presentation)

2010-201420112012

2012

2012

2012

2013

2013

2014

2014

2508

20

10

12

40

50

20

16

12

(Inter)national conferences- 12th European Biological Rhythms Society

(ERBS) meeting Oxford- Society for Research on Biological Rhythms

(SRBR) meeting Big Sky, Montana (oral presentation)

2011

2014

40

50

2. TeachingYear Workload

(Hours)Other- Supervising Junior Science Program- Supervising Bachelor thesis Hogeschool

Utrecht

20122012-2013

80100


139

8

Dankwoord

Dankwoord

Na bijna zes jaar lang hard werken is mijn proefschrift af! En nu is het tijd om

iedereen die daaraan heeft bijgedragen te bedanken.

Allereerst mijn promotoren. Bert, je introduceerde me in de wereld van de

chronobiologie en op een beperkte afstand was je altijd betrokken bij mijn

project. Dankjewel voor je vertrouwen en geduld dat hiermee gepaard ging. Harry,

dankjewel voor je kritische vragen en soms pittige discussies, die ervoor zorgden

dat ik uiteindelijk in de juiste richting weer verder kon met mijn onderzoek.

Wendy, jouw deur stond altijd open. In eerste instantie om samen tijdschema’s te

tekenen en een weg te vinden in de terminologie van de chronobiologie. Later

hielp je om de focus en planning van alle verschillende studies in het oog te

houden.

Mijn praktische steun en toeverlaat, Conny. Zonder jouw hulp bij het uitvoeren

van alle experimenten, en het draaien van vele qPCRs en ELISAs was dit boekje

niet mogelijk geweest. Zelfs in de nachtelijke uurtjes hielp je mee samples te

verzamelen. Vaak was dit naast nuttig ook erg gezellig!

Linda, ongeveer halverwege mijn promotietraject werd jij toegevoegd aan onze

dag-nachtclub. Naast je kritische vragen over de inhoud van mijn onderzoek,

kletsten we tijdens onze lunchwandelingen over nog veel meer. Naast een fijne

wetenschappelijke sparringspartner, was en ben je altijd een luisterend oor,

dankjewel hiervoor!

Professor Roenneberg, dear Till, thank you for the inspiring and fruitful discussions

we had via skype and during our visit in Priel. Your input is valued a lot.

Lidewij, bedankt voor jouw bijdrage aan mijn onderzoek als stagiaire op het dag-

nachtproject.

Edwin, Liset en Paul, mede met jullie inzet heb ik nog een mooie appendix aan

mijn proefschrift kunnen toevoegen. Mirjam, dankjewel dat je me niet kopje onder

hebt laten gaan in deze transcriptomics zee van data.


140

Chapter 8

Mijn kamergenootjes, Sjors, Ilse, Myrtho, Josh, Linda en Charlotte, bedankt voor

jullie gezelligheid. Maar ook voor het meedenken, advies geven en luisterend

oor. En ook alle andere mede-aio’s, Anne, Peter, Sanne, Marja, Mirjam, Joantine en

Sander, samen in hetzelfde schuitje maakt het wel zo gezellig.

Dan zijn er nog vele collega’s om te bedanken. GBO en later GZB, bedankt voor de

ontspanningsmomentjes tijdens de koffie, lunch of een borrel. En ook de collega’s

in Rotterdam, bedankt voor jullie gastvrijheid op de momenten dat ik daar was voor

mijn studies. Speciaal bedankt, Stefanie en Sylvia, jullie hebben ervoor gezorgd dat

het mogelijk was om op twee plekken tegelijk studies uit te voeren.

Naast werk en wetenschap was er ook tijd voor andere zaken. Jochem,

Hedwig, Thijs en Sanne, ontspanning en ontsnapping zijn de code-woorden.

Met zoveel doctors in de club moet het lukken om nog meer snelste

tijden op de klok te zetten. En Hedwig, fijn dat je mijn paranimf wil zijn.

Anne, wat leuk dat we nu wederom collega’s zijn. Jij brengt altijd gezelligheid mee

en ik ben blij dat je mijn paranimf wil zijn.

Nicky, Marja, Marike en Karin, Nijmegen vrienden van het eerste uur. Inmiddels niet

meer wekelijks sporten en samen eten, maar we blijven momentjes zoeken om

gezellig bij te kletsen.

En alle andere vrienden, bedankt voor jullie interesse in mijn onderzoek en

gezelligheid tijdens borrels, etentjes en feestjes.

Arjan, Ria, Alwin en Lysette, jullie kennen me ondertussen al een lange tijd. Wat fijn

dat jullie nu ook deze mijlpaal mee mogen beleven. Lysette, heel erg bedankt voor

de mooie omslag van dit proefschrift, zo is het helemaal af.

Karin, Arnoud, Jeroen, Cora en Arjan, wat fijn dat jullie altijd interesse hadden in

mijn onderzoek.

Karin, zusje, wat fijn dat we zoveel leuke dingen samen kunnen doen. Ook samen

met Alfred. Deze momenten vind ik ontzettend waardevol.


141

8

Dankwoord

Lieve Mam en Arie, jullie hebben altijd mijn nieuwsgierigheid en wil om te

ontdekken gestimuleerd. Jullie vertrouwen in mijn kunnen heeft me gesteund en

zeker geholpen om mijn onderzoek tot een goed eind te brengen. Dankjulliewel!

Lieve Stefan, dankjewel voor geduld en vertrouwen de afgelopen jaren. Jij zorgt

ervoor dat ik elke dag het beste uit mezelf kan halen. Bij jou kan ik mezelf zijn en

samen is het leven leuker en gezelliger. Ik hou van jou!


142

Chapter 8

ReferencesAKERSTEDT, T. 2003. Shift work and disturbed sleep/wakefulness. Occup Med (Lond), 53, 89-94.

AKERSTEDT, T., KNUTSSON, A., NARUSYTE, J., SVEDBERG, P., KECKLUND, G. & ALEXANDERSON, K. 2015. Night work and breast cancer in women: a Swedish cohort study. BMJ Open, 5, e008127.

AKERSTEDT, T. & WRIGHT, K. P., JR. 2009. Sleep Loss and Fatigue in Shift Work and Shift Work Disorder. Sleep Med Clin, 4, 257-271.

AKHTAR, R. A., REDDY, A. B., MAYWOOD, E. S., CLAYTON, J. D., KING, V. M., SMITH, A. G., GANT, T. W., HASTINGS, M. H. & KYRIACOU, C. P. 2002. Circadian cycling of the mouse liver transcriptome, as revealed by cDNA microarray, is driven by the suprachiasmatic nucleus. Curr Biol, 12, 540-50.

ANDERSON, L. E., MORRIS, J. E., SASSER, L. B. & STEVENS, R. G. 2000. Effect of constant light on DMBA mammary tumorigenesis in rats. Cancer Lett, 148, 121-6.

ANDERSON, N. L., POLANSKI, M., PIEPER, R., GATLIN, T., TIRUMALAI, R. S., CONRADS, T. P., VEENSTRA, T. D., ADKINS, J. N., POUNDS, J. G., FAGAN, R. & LOBLEY, A. 2004. The human plasma proteome: a nonredundant list developed by combination of four separate sources. Mol.Cell Proteomics., 3, 311-326.

ANG, J. E., REVELL, V., MANN, A., MANTELE, S., OTWAY, D. T., JOHNSTON, J. D., THUMSER, A. E., SKENE, D. J. & RAYNAUD, F. 2012. Identification of human plasma metabolites exhibiting time-of-day variation using an untargeted liquid chromatography-mass spectrometry metabolomic approach. Chronobiol Int, 29, 868-81.

ANISIMOV, V. N., BATURIN, D. A., POPOVICH, I. G., ZABEZHINSKI, M. A., MANTON, K. G., SEMENCHENKO, A. V. & YASHIN, A. I. 2004. Effect of exposure to light-at-night on life span and spontaneous carcinogenesis in female CBA mice. Int J Cancer, 111, 475-9.

ARCHER, S. N., LAING, E. E., MOLLER-LEVET, C. S., VAN DER VEEN, D. R., BUCCA, G., LAZAR, A. S., SANTHI, N., SLAK, A., KABILJO, R., VON SCHANTZ, M., SMITH, C. P. & DIJK, D. J. 2014a. Mistimed sleep disrupts circadian regulation of the human transcriptome. Proc Natl Acad Sci U S A.

ARCHER, S. N., LAING, E. E., MOLLER-LEVET, C. S., VAN DER VEEN, D. R., BUCCA, G., LAZAR, A. S., SANTHI, N., SLAK, A., KABILJO, R., VON SCHANTZ, M., SMITH, C. P. & DIJK, D. J. 2014b. Mistimed sleep disrupts circadian regulation of the human transcriptome. Proc Natl Acad Sci U S A, 111, E682-91.

ARNOLD, M. & LANGHANS, W. 2010. Effects of anesthesia and blood sampling techniques on plasma metabolites and corticosterone in the rat. Physiol Behav, 99, 592-8.

ASCHOFF, J., FATRANSKA, M., GIEDKE, H., DOERR, P., STAMM, D. & WISSER, H. 1971. Human circadian rhythms in continuous darkness: entrainment by social cues. Science, 171, 213-5.

AUBERT, C., JANIAUD, P. & LECALVEZ, J. 1980. Effect of pinealectomy and melatonin on mammary tumor growth in Sprague-Dawley rats under different conditions of lighting. J Neural Transm, 47, 121-30.

BAILEY, S. M., UDOH, U. S. & YOUNG, M. E. 2014. Circadian regulation of metabolism. J Endocrinol, 222, R75-96.

BALMAIN, A. & HARRIS, C. C. 2000. Carcinogenesis in mouse and human cells: parallels and paradoxes. Carcinogenesis, 21, 371-7.

BARCLAY, J. L., HUSSE, J., BODE, B., NAUJOKAT, N., MEYER-KOVAC, J., SCHMID, S. M., LEHNERT, H. & OSTER, H. 2012. Circadian desynchrony promotes metabolic disruption in a mouse model of shiftwork. PLoS One, 7, e37150.

BARTON, J., FOLKARD, S., SMITH, L. & POOLE, C. J. 1994. Effects on health of a change from a delaying to an advancing shift system. Occup Environ Med, 51, 749-55.

BELL-PEDERSEN, D., CASSONE, V. M., EARNEST, D. J., GOLDEN, S. S., HARDIN, P. E., THOMAS, T. L. & ZORAN, M. J. 2005. Circadian rhythms from multiple oscillators: lessons from diverse organisms. Nat Rev Genet, 6, 544-56.

BENJAMINI, Y. & HOCHBERG, Y. 1995. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological), 57, 289-300.

BERENBLUM, I. (ed.) 1975. Sequential Aspect of Chemical Carcinogenesis: Skin, New York: Plenum Press.

BERSON, D. M., DUNN, F. A. & TAKAO, M. 2002. Phototransduction by retinal ganglion cells that set the circadian clock. Science, 295, 1070-3.

BIANCHINI, F., KAAKS, R. & VAINIO, H. 2002. Overweight, obesity, and cancer risk. Lancet Oncol, 3, 565-74.

BIRKETVEDT, G. S., GELIEBTER, A., KRISTIANSEN, I., FIRGENSCHAU, Y., GOLL, R. & FLORHOLMEN, J. R. 2012. Diurnal secretion of ghrelin, growth hormone, insulin binding proteins, and prolactin in normal weight and overweight subjects with and without the night eating syndrome. Appetite, 59, 688-92.


143

8

References

BLASK, D. E., BRAINARD, G. C., DAUCHY, R. T., HANIFIN, J. P., DAVIDSON, L. K., KRAUSE, J. A., SAUER, L. A., RIVERA-BERMUDEZ, M. A., DUBOCOVICH, M. L., JASSER, S. A., LYNCH, D. T., ROLLAG, M. D. & ZALATAN, F. 2005. Melatonin-depleted blood from premenopausal women exposed to light at night stimulates growth of human breast cancer xenografts in nude rats. Cancer Res, 65, 11174-84.

BLASK, D. E., DAUCHY, R. T., DAUCHY, E. M., MAO, L., HILL, S. M., GREENE, M. W., BELANCIO, V. P., SAUER, L. A. & DAVIDSON, L. 2014. Light exposure at night disrupts host/cancer circadian regulatory dynamics: impact on the Warburg effect, lipid signaling and tumor growth prevention. PLoS One, 9, e102776.

BLASK, D. E., DAUCHY, R. T., SAUER, L. A., KRAUSE, J. A. & BRAINARD, G. C. 2003. Growth and fatty acid metabolism of human breast cancer (MCF-7) xenografts in nude rats: impact of constant light-induced nocturnal melatonin suppression. Breast Cancer Res Treat, 79, 313-20.

BLASK, D. E., PELLETIER, D. B., HILL, S. M., LEMUS-WILSON, A., GROSSO, D. S., WILSON, S. T. & WISE, M. E. 1991. Pineal melatonin inhibition of tumor promotion in the N-nitroso-N-methylurea model of mammary carcinogenesis: potential involvement of antiestrogenic mechanisms in vivo. J Cancer Res Clin Oncol, 117, 526-32.

BOGGILD, H. & JEPPESEN, H. J. 2001. Intervention in shift scheduling and changes in biomarkers of heart disease in hospital wards. Scand J Work Environ Health, 27, 87-96.

BÖHM, D., KELLER, K., WEHRWEIN, N., LEBRECHT, A., SCHMIDT, M., KOLBL, H. & GRUS, F. H. 2011. Serum proteome profiling of primary breast cancer indicates a specific biomarker profile. Oncol Rep, 26, 1051-6.

BOIVIN, D. B., BOUDREAU, P., JAMES, F. O. & KIN, N. M. 2012. Photic resetting in night-shift work: impact on nurses’ sleep. Chronobiol Int, 29, 619-28.

BOQUET, A., CRUZ, C., NESTHUS, T., DETWILER, C., KNECHT, W. & HOLCOMB, K. 2004. Clockwise and counterclockwise rotating shifts: effects on temperature and neuroendocrine measures. Aviat Space Environ Med, 75, 898-904.

BORRESEN-DALE, A. L. 2003. TP53 and breast cancer. Hum Mutat, 21, 292-300.

BOVBJERG, D. H. 2003. Circadian disruption and cancer: sleep and immune regulation. Brain Behav Immun, 17 Suppl 1, S48-50.

BOZEK, K., RELOGIO, A., KIELBASA, S. M., HEINE, M., DAME, C., KRAMER, A. & HERZEL, H. 2009. Regulation of clock-controlled genes in mammals. PLoS One, 4, e4882.

BRAMBILLA, D. J., MATSUMOTO, A. M., ARAUJO, A. B. & MCKINLAY, J. B. 2009. The effect of diurnal variation on clinical measurement of serum testosterone and other sex hormone levels in men. J Clin Endocrinol Metab, 94, 907-13.

BREIMAN, L. 2001. Random Forests. Machine Learning, 45, 5-32.

BUDNICK, L. D., LERMAN, S. E. & NICOLICH, M. J. 1995. An evaluation of scheduled bright light and darkness on rotating shiftworkers: trial and limitations. Am J Ind Med, 27, 771-82.

CAHILL, G. M., HURD, M. W. & BATCHELOR, M. M. 1998. Circadian rhythmicity in the locomotor activity of larval zebrafish. Neuroreport, 9, 3445-9.

CAPIZZI, E., GRUPPIONI, E., GRIGIONI, A. D., GABUSI, E., GRASSIGLI, A., GRIGIONI, W. F. & FIORENTINO, M. 2008. Real time RT-PCR approach for the evaluation of ERBB2 overexpression in breast cancer archival samples: a comparative study with FISH, SISH, and immunohistochemistry. Diagn Mol Pathol, 17, 220-6.

CARNEIRO, B. T. & ARAUJO, J. F. 2012. Food entrainment: major and recent findings. Front Behav Neurosci, 6, 83.

CBS. 2010. CBS Statline Beroepsbevolking; regeling werktijden [Online]. Available: http://statline.cbs.nl/StatWeb/publication/?DM=SLNL&PA=71959ned&D1=2-6&D2=0&D3=a&D4=a&HDR=T&STB=G1,G2,G3&VW=T.

CHAVES, I., NIJMAN, R. M., BIERNAT, M. A., BAJEK, M. I., BRAND, K., DA SILVA, A. C., SAITO, S., YAGITA, K., EKER, A. P. & VAN DER HORST, G. T. 2011. The Potorous CPD photolyase rescues a cryptochrome-deficient mammalian circadian clock. PLoS One, 6, e23447.

CHUA, E. C., SHUI, G., LEE, I. T., LAU, P., TAN, L. C., YEO, S. C., LAM, B. D., BULCHAND, S., SUMMERS, S. A., PUVANENDRAN, K., ROZEN, S. G., WENK, M. R. & GOOLEY, J. J. 2013. Extensive diversity in circadian regulation of plasma lipids and evidence for different circadian metabolic phenotypes in humans. Proc Natl Acad Sci U S A, 110, 14468-73.

CLIMENT, J., PEREZ-LOSADA, J., QUIGLEY, D. A., KIM, I. J., DELROSARIO, R., JEN, K. Y., BOSCH, A., LLUCH, A., MAO, J. H. & BALMAIN, A. 2010. Deletion of the PER3 gene on chromosome 1p36 in recurrent ER-positive breast cancer. J Clin Oncol, 28, 3770-8.

COLES, C., CONDIE, A., CHETTY, U., STEEL, C. M., EVANS, H. J. & PROSSER, J. 1992. p53 mutations in breast cancer. Cancer Res, 52, 5291-8.


144

Chapter 8

COOMANS, C. P., VAN DEN BERG, S. A., HOUBEN, T., VAN KLINKEN, J. B., VAN DEN BERG, R., PRONK, A. C., HAVEKES, L. M., ROMIJN, J. A., VAN DIJK, K. W., BIERMASZ, N. R. & MEIJER, J. H. 2013. Detrimental effects of constant light exposure and high-fat diet on circadian energy metabolism and insulin sensitivity. FASEB J, 27, 1721-32.

COS, S., MEDIAVILLA, D., MARTINEZ-CAMPA, C., GONZALEZ, A., ALONSO-GONZALEZ, C. & SANCHEZ-BARCELO, E. J. 2006. Exposure to light-at-night increases the growth of DMBA-induced mammary adenocarcinomas in rats. Cancer Lett, 235, 266-71.

CULHANE, A. C., SCHWARZL, T., SULTANA, R., PICARD, K. C., PICARD, S. C., LU, T. H., FRANKLIN, K. R., FRENCH, S. J., PAPENHAUSEN, G., CORRELL, M. & QUACKENBUSH, J. 2010. GeneSigDB--a curated database of gene expression signatures. Nucleic Acids Res, 38, D716-25.

CZEISLER, C. A., DUFFY, J. F., SHANAHAN, T. L., BROWN, E. N., MITCHELL, J. F., RIMMER, D. W., RONDA, J. M., SILVA, E. J., ALLAN, J. S., EMENS, J. S., DIJK, D. J. & KRONAUER, R. E. 1999. Stability, precision, and near-24-hour period of the human circadian pacemaker. Science, 284, 2177-81.

DAAN, S., MERROW, M. & ROENNEBERG, T. 2002. External time--internal time. J Biol Rhythms, 17, 107-9.

DALLMANN, R., VIOLA, A. U., TAROKH, L., CAJOCHEN, C. & BROWN, S. A. 2012. The human circadian metabolome. Proc Natl Acad Sci U S A, 109, 2625-9.

DANAEI, G., FINUCANE, M. M., LU, Y., SINGH, G. M., COWAN, M. J., PACIOREK, C. J., LIN, J. K., FARZADFAR, F., KHANG, Y. H., STEVENS, G. A., RAO, M., ALI, M. K., RILEY, L. M., ROBINSON, C. A. & EZZATI, M. 2011. National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2.7 million participants. Lancet, 378, 31-40.

DAVIDSON, A. J., SELLIX, M. T., DANIEL, J., YAMAZAKI, S., MENAKER, M. & BLOCK, G. D. 2006. Chronic jet-lag increases mortality in aged mice. Curr Biol, 16, R914-6.

DAVIES, S. K., ANG, J. E., REVELL, V. L., HOLMES, B., MANN, A., ROBERTSON, F. P., CUI, N., MIDDLETON, B., ACKERMANN, K., KAYSER, M., THUMSER, A. E., RAYNAUD, F. I. & SKENE, D. J. 2014. Effect of sleep deprivation on the human metabolome. Proc Natl Acad Sci U S A, 111, 10761-6.

DAVIS, S. & MIRICK, D. K. 2006. Circadian disruption, shift work and the risk of cancer: a summary of the evidence and studies in Seattle. Cancer Causes Control, 17, 539-45.

DEFILIPPIS, R. A., CHANG, H., DUMONT, N., RABBAN, J. T., CHEN, Y. Y., FONTENAY, G. V., BERMAN, H. K., GAUTHIER, M. L., ZHAO, J., HU, D., MARX, J. J., TJOE, J. A., ZIV, E., FEBBRAIO, M., KERLIKOWSKE, K., PARVIN, B. & TLSTY, T. D. 2012. CD36 repression activates a multicellular stromal program shared by high mammographic density and tumor tissues. Cancer Discov, 2, 826-39.

DEGENHARDT, T., SARAMAKI, A., MALINEN, M., RIECK, M., VAISANEN, S., HUOTARI, A., HERZIG, K. H., MULLER, R. & CARLBERG, C. 2007. Three members of the human pyruvate dehydrogenase kinase gene family are direct targets of the peroxisome proliferator-activated receptor beta/delta. J Mol Biol, 372, 341-55.

DERKSEN, P. W., BRAUMULLER, T. M., VAN DER BURG, E., HORNSVELD, M., MESMAN, E., WESSELING, J., KRIMPENFORT, P. & JONKERS, J. 2011. Mammary-specific inactivation of E-cadherin and p53 impairs functional gland development and leads to pleomorphic invasive lobular carcinoma in mice. Dis Model Mech, 4, 347-58.

DIBNER, C., SCHIBLER, U. & ALBRECHT, U. 2010. The mammalian circadian timing system: organization and coordination of central and peripheral clocks. Annu Rev Physiol, 72, 517-49.

DIVER, M. J., IMTIAZ, K. E., AHMAD, A. M., VORA, J. P. & FRASER, W. D. 2003. Diurnal rhythms of serum total, free and bioavailable testosterone and of SHBG in middle-aged men compared with those in young men. Clin Endocrinol (Oxf ), 58, 710-7.

DOLLE, M. E., MARTUS, H. J., GOSSEN, J. A., BOERRIGTER, M. E. & VIJG, J. 1996. Evaluation of a plasmid-based transgenic mouse model for detecting in vivo mutations. Mutagenesis, 11, 111-8.

DUFFY, J. F., KRONAUER, R. E. & CZEISLER, C. A. 1996. Phase-shifting human circadian rhythms: influence of sleep timing, social contact and light exposure. J Physiol, 495 ( Pt 1), 289-97.

ECKEL-MAHAN, K. L., PATEL, V. R., DE MATEO, S., OROZCO-SOLIS, R., CEGLIA, N. J., SAHAR, S., DILAG-PENILLA, S. A., DYAR, K. A., BALDI, P. & SASSONE-CORSI, P. 2013. Reprogramming of the circadian clock by nutritional challenge. Cell, 155, 1464-78.

EDRN. Available: http://edrn.nci.nih.gov/.

EIN-DOR, L., KELA, I., GETZ, G., GIVOL, D. & DOMANY, E. 2005. Outcome signature genes in breast cancer: is there a unique set? Bioinformatics, 21, 171-8.


145

8

References

EMENS, J. S., YUHAS, K., ROUGH, J., KOCHAR, N., PETERS, D. & LEWY, A. J. 2009. Phase angle of entrainment in morning- and evening-types under naturalistic conditions. Chronobiol Int, 26, 474-93.

EUROFOUND 2012. Fifth European Working Conditions Survey.

FARRAH, T., DEUTSCH, E. W., OMENN, G. S., CAMPBELL, D. S., SUN, Z., BLETZ, J. A., MALLICK, P., KATZ, J. E., MALMSTROM, J., OSSOLA, R., WATTS, J. D., LIN, B., ZHANG, H., MORITZ, R. L. & AEBERSOLD, R. 2011. A high-confidence human plasma proteome reference set with estimated concentrations in PeptideAtlas. Mol Cell Proteomics, 10, M110 006353.

FEBBRAIO, M., HAJJAR, D. P. & SILVERSTEIN, R. L. 2001. CD36: a class B scavenger receptor involved in angiogenesis, atherosclerosis, inflammation, and lipid metabolism. J Clin Invest, 108, 785-91.

FERLAY, J., SHIN, H. R., BRAY, F., FORMAN, D., MATHERS, C. & PARKIN, D. M. 2010. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer, 127, 2893-917.

FERLAY, J., SOERJOMATARAM, I., DIKSHIT, R., ESER, S., MATHERS, C., REBELO, M., PARKIN, D. M., FORMAN, D. & BRAY, F. 2015. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer, 136, E359-86.

FILIPSKI, E., DELAUNAY, F., KING, V. M., WU, M. W., CLAUSTRAT, B., GRECHEZ-CASSIAU, A., GUETTIER, C., HASTINGS, M. H. & FRANCIS, L. 2004. Effects of chronic jet lag on tumor progression in mice. Cancer Res, 64, 7879-85.

FILIPSKI, E., INNOMINATO, P. F., WU, M., LI, X. M., IACOBELLI, S., XIAN, L. J. & LEVI, F. 2005. Effects of light and food schedules on liver and tumor molecular clocks in mice. J Natl Cancer Inst, 97, 507-17.

FILIPSKI, E., SUBRAMANIAN, P., CARRIERE, J., GUETTIER, C., BARBASON, H. & LEVI, F. 2009. Circadian disruption accelerates liver carcinogenesis in mice. Mutat Res, 680, 95-105.

FISHER, S. P., GODINHO, S. I., POTHECARY, C. A., HANKINS, M. W., FOSTER, R. G. & PEIRSON, S. N. 2012. Rapid assessment of sleep-wake behavior in mice. J Biol Rhythms, 27, 48-58.

FOLKMAN, J. 1985. Tumor angiogenesis. Adv Cancer Res, 43, 175-203.

FONKEN, L. K., WORKMAN, J. L., WALTON, J. C., WEIL, Z. M., MORRIS, J. S., HAIM, A. & NELSON, R. J. 2010. Light at night increases body mass by shifting the time of food intake. Proc Natl Acad Sci U S A, 107, 18664-9.

FRITSCHI, L., ERREN, T. C., GLASS, D. C., GIRSCHIK, J., THOMSON, A. K., SAUNDERS, C., BOYLE, T., EL-ZAEMEY, S., ROGERS, P., PETERS, S., SLEVIN, T., D’ORSOGNA, A., DE VOCHT, F., VERMEULEN, R. & HEYWORTH, J. S. 2013. The association between different night shiftwork factors and breast cancer: a case-control study. Br J Cancer, 109, 2472-80.

FRITSCHI, L., GLASS, D. C., HEYWORTH, J. S., ARONSON, K., GIRSCHIK, J., BOYLE, T., GRUNDY, A. & ERREN, T. C. 2011. Hypotheses for mechanisms linking shiftwork and cancer. Med Hypotheses, 77, 430-6.

FU, L., PELICANO, H., LIU, J., HUANG, P. & LEE, C. 2002. The circadian gene Period2 plays an important role in tumor suppression and DNA damage response in vivo. Cell, 111, 41-50.

FUJIMOTO, V. Y., CLIFTON, D. K., COHEN, N. L. & SOULES, M. R. 1990. Variability of serum prolactin and progesterone levels in normal women: the relevance of single hormone measurements in the clinical setting. Obstet Gynecol, 76, 71-8.

FUJINO, Y., ISO, H., TAMAKOSHI, A., INABA, Y., KOIZUMI, A., KUBO, T., YOSHIMURA, T. & JAPANESE COLLABORATIVE COHORT STUDY, G. 2006. A prospective cohort study of shift work and risk of ischemic heart disease in Japanese male workers. Am J Epidemiol, 164, 128-35.

GAMBLE, K. L., BERRY, R., FRANK, S. J. & YOUNG, M. E. 2014. Circadian clock control of endocrine factors. Nat Rev Endocrinol, 10, 466-75.

GARLAND, C. F. & GARLAND, F. C. 2006. Do sunlight and vitamin D reduce the likelihood of colon cancer? Int J Epidemiol, 35, 217-20.

GIANNI, L., DAFNI, U., GELBER, R. D., AZAMBUJA, E., MUEHLBAUER, S., GOLDHIRSCH, A., UNTCH, M., SMITH, I., BASELGA, J., JACKISCH, C., CAMERON, D., MANO, M., PEDRINI, J. L., VERONESI, A., MENDIOLA, C., PLUZANSKA, A., SEMIGLAZOV, V., VRDOLJAK, E., ECKART, M. J., SHEN, Z., SKIADOPOULOS, G., PROCTER, M., PRITCHARD, K. I., PICCART-GEBHART, M. J. & BELL, R. 2011. Treatment with trastuzumab for 1 year after adjuvant chemotherapy in patients with HER2-positive early breast cancer: a 4-year follow-up of a randomised controlled trial. Lancet Oncol, 12, 236-44.

GOOLEY, J. J. & CHUA, E. C. 2014. Diurnal regulation of lipid metabolism and applications of circadian lipidomics. J Genet Genomics, 41, 231-50.

GRUNDY, A., SANCHEZ, M., RICHARDSON, H., TRANMER, J., BORUGIAN, M., GRAHAM, C. H. & ARONSON, K. J. 2009. Light intensity exposure, sleep duration, physical activity, and biomarkers of melatonin among rotating shift nurses. Chronobiol Int, 26, 1443-61.


146

Chapter 8

GU, F., HAN, J., LADEN, F., PAN, A., CAPORASO, N. E., STAMPFER, M. J., KAWACHI, I., REXRODE, K. M., WILLETT, W. C., HANKINSON, S. E., SPEIZER, F. E. & SCHERNHAMMER, E. S. 2015. Total and cause-specific mortality of U.S. nurses working rotating night shifts. Am J Prev Med, 48, 241-52.

GUTMAN, H., POLLOCK, R. E., JANJAN, N. A. & JOHNSTON, D. A. 1995. Biologic distinctions and therapeutic implications of sarcomatoid metaplasia of epithelial carcinoma of the breast. J Am Coll Surg, 180, 193-9.

HADJI, P., KAUKA, A., ZILLER, M., BIRKHOLZ, K., BAIER, M., MUTH, M. & KANN, P. 2014. Effect of adjuvant endocrine therapy on hormonal levels in premenopausal women with breast cancer: the ProBONE II study. Breast Cancer Res Treat, 144, 343-51.

HAMILTON, T. 1969. Influence of environmental light and melatonin upon mammary tumour induction. Br J Surg, 56, 764-6.

HANDBERG, A., LEVIN, K., HOJLUND, K. & BECK-NIELSEN, H. 2006. Identification of the oxidized low-density lipoprotein scavenger receptor CD36 in plasma: a novel marker of insulin resistance. Circulation, 114, 1169-76.

HARDELAND, R., PANDI-PERUMAL, S. R. & CARDINALI, D. P. 2006. Melatonin. Int J Biochem Cell Biol, 38, 313-6.

HARMA, M., TARJA, H., IRJA, K., MIKAEL, S., JUSSI, V., ANNE, B. & PERTTI, M. 2006. A controlled intervention study on the effects of a very rapidly forward rotating shift system on sleep-wakefulness and well-being among young and elderly shift workers. Int J Psychophysiol, 59, 70-9.

HASTINGS, M. H., DUFFIELD, G. E., SMITH, E. J., MAYWOOD, E. S. & EBLING, F. J. 1998. Entrainment of the circadian system of mammals by nonphotic cues. Chronobiol Int, 15, 425-45.

HE, C., ANAND, S. T., EBELL, M. H., VENA, J. E. & ROBB, S. W. 2014. Circadian disrupting exposures and breast cancer risk: a meta-analysis. Int Arch Occup Environ Health.

HEALTH COUNCIL OF THE NETHERLANDS. Shift work and health risks: possibilities for prevention. The Hague: Health Council of the Netherlands, 2015, publication no. 2015/25

HIRSCHFELD, U., MORENO-REYES, R., AKSEKI, E., L’HERMITE-BALERIAUX, M., LEPROULT, R., COPINSCHI, G. & VAN CAUTER, E. 1996. Progressive elevation of plasma thyrotropin during adaptation to simulated jet lag: effects of treatment with bright light or zolpidem. J Clin Endocrinol Metab, 81, 3270-7.

HUANG, D. W., SHERMAN, B. T. & LEMPICKI, R. A. 2009. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc., 4, 44-57.

HUGHES, M. E., HOGENESCH, J. B. & KORNACKER, K. 2010. JTK_CYCLE: an efficient nonparametric algorithm for detecting rhythmic components in genome-scale data sets. J Biol Rhythms, 25, 372-80.

HURD, M. W., DEBRUYNE, J., STRAUME, M. & CAHILL, G. M. 1998. Circadian rhythms of locomotor activity in zebrafish. Physiol Behav, 65, 465-72.

IARC 2010. Monograph Volume 98: Painting, firefighting, and shift work / IARC Working Group on the Evaluation of Carcinogenic Risks to Humans.

IJAZ, S., VERBEEK, J., SEIDLER, A., LINDBOHM, M. L., OJAJARVI, A., ORSINI, N., COSTA, G. & NEUVONEN, K. 2013. Night-shift work and breast cancer--a systematic review and meta-analysis. Scand J Work Environ Health, 39, 431-47.

ITOH, H., WENG, Z., SAITO, H., OGAWA, Y., NAKAYAMA, K., HASEGAWA-OHIRA, M., MORIMOTO, K., MAKI, S. & TAKAHASHI, M. 2011. Association between night-shift work and serum 25-hydroxyvitamin D levels in Japanese male indoor workers: a cross-sectional study. Ind Health, 49, 658-62.

ITTMANN, M., HUANG, J., RADAELLI, E., MARTIN, P., SIGNORETTI, S., SULLIVAN, R., SIMONS, B. W., WARD, J. M., ROBINSON, B. D., CHU, G. C., LODA, M., THOMAS, G., BOROWSKY, A. & CARDIFF, R. D. 2013. Animal models of human prostate cancer: the consensus report of the New York meeting of the Mouse Models of Human Cancers Consortium Prostate Pathology Committee. Cancer Res, 73, 2718-36.

IWAMOTO, A., KAWAI, M., FURUSE, M. & YASUO, S. 2014. Effects of chronic jet lag on the central and peripheral circadian clocks in CBA/N mice. Chronobiol Int, 31, 189-98.

JAMES, F. O., BOIVIN, D. B., CHARBONNEAU, S., BELANGER, V. & CERMAKIAN, N. 2007. Expression of clock genes in human peripheral blood mononuclear cells throughout the sleep/wake and circadian cycles. Chronobiol Int, 24, 1009-34.

JEOUNG, N. H. & HARRIS, R. A. 2010. Role of pyruvate dehydrogenase kinase 4 in regulation of blood glucose levels. Korean Diabetes J, 34, 274-83.

JIA, Y., LU, Y., WU, K., LIN, Q., SHEN, W., ZHU, M., HUANG, S. & CHEN, J. 2013. Does night work increase the risk of breast cancer? A systematic review and meta-analysis of epidemiological studies. Cancer Epidemiol, 37, 197-206.


147

8

References

JIANG, J., BOYLE, L. J., MIKUS, C. R., OBERLIN, D. J., FLETCHER, J. A., THYFAULT, J. P. & HINTON, P. S. 2014. The effects of improved metabolic risk factors on bone turnover markers after 12 weeks of simvastatin treatment with or without exercise. Metabolism, 63, 1398-408.

JUDA, M., VETTER, C. & ROENNEBERG, T. 2013a. Chronotype modulates sleep duration, sleep quality, and social jet lag in shift-workers. J Biol Rhythms, 28, 141-51.

JUDA, M., VETTER, C. & ROENNEBERG, T. 2013b. The Munich ChronoType Questionnaire for Shift-Workers (MCTQShift). J Biol Rhythms, 28, 130-40.

JUNG, C. M., KHALSA, S. B., SCHEER, F. A., CAJOCHEN, C., LOCKLEY, S. W., CZEISLER, C. A. & WRIGHT, K. P., JR. 2010. Acute effects of bright light exposure on cortisol levels. J Biol Rhythms, 25, 208-16.

KAMDAR, B. B., TERGAS, A. I., MATEEN, F. J., BHAYANI, N. H. & OH, J. 2013. Night-shift work and risk of breast cancer: a systematic review and meta-analysis. Breast Cancer Res Treat, 138, 291-301.

KARIM, B. O. & HUSO, D. L. 2013. Mouse models for colorectal cancer. Am J Cancer Res, 3, 240-50.

KATZ, L. E., DELEON, D. D., ZHAO, H. & JAWAD, A. F. 2002. Free and total insulin-like growth factor (IGF)-I levels decline during fasting: relationships with insulin and IGF-binding protein-1. J Clin Endocrinol Metab, 87, 2978-83.

KELLY-SPRATT, K. S., KASARDA, A. E., IGRA, M. & KEMP, C. J. 2008. A mouse model repository for cancer biomarker discovery. J Proteome Res, 7, 3613-8.

KHAETSKI, I. K. 1965. Effect of hypothalamo-pituitary lesions induced by constant illumination on development of induced mammary tumors in rats. Vopr Exp Oncol, 1, 87–93.

KIESSLING, S., EICHELE, G. & OSTER, H. 2010. Adrenal glucocorticoids have a key role in circadian resynchronization in a mouse model of jet lag. J Clin Invest, 120, 2600-9.

KIMLIN, M. G. & TENKATE, T. D. 2007. Occupational exposure to ultraviolet radiation: the duality dilemma. Rev Environ Health, 22, 1-37.

KLEIN, A., WESSEL, R., GRAESSMANN, M., JURGENS, M., PETERSEN, I., SCHMUTZLER, R., NIEDERACHER, D., ARNOLD, N., MEINDL, A., SCHERNECK, S., SEITZ, S. & GRAESSMANN, A. 2007. Comparison of gene expression data from human and mouse breast cancers: identification of a conserved breast tumor gene set. Int J Cancer, 121, 683-8.

KLINGMAN, K. M., MARSH, E. E., KLERMAN, E. B., ANDERSON, E. J. & HALL, J. E. 2011. Absence of circadian rhythms of gonadotropin secretion in women. J Clin Endocrinol Metab, 96, 1456-61.

KNAUTH, P. 1996. Designing better shift systems. Appl Ergon, 27, 39-44.

KNUTSSON, A. & BOGGILD, H. 2010. Gastrointestinal disorders among shift workers. Scand J Work Environ Health, 36, 85-95.

KO, C. H. & TAKAHASHI, J. S. 2006. Molecular components of the mammalian circadian clock. Hum Mol Genet, 15 Spec No 2, R271-7.

KOTHARI, L. S., SHAH, P. N. & MHATRE, M. C. 1984. Pineal ablation in varying photoperiods and the incidence of 9,10-dimethyl-1,2-benzanthracene induced mammary cancer in rats. Cancer Lett, 22, 99-102.

KRETSCHMER, C., STERNER-KOCK, A., SIEDENTOPF, F., SCHOENEGG, W., SCHLAG, P. M. & KEMMNER, W. 2011. Identification of early molecular markers for breast cancer. Mol Cancer, 10, 15.

KUBO, T., OYAMA, I., NAKAMURA, T., SHIRANE, K., OTSUKA, H., KUNIMOTO, M., KADOWAKI, K., MARUYAMA, T., OTOMO, H., FUJINO, Y., MATSUMOTO, T. & MATSUDA, S. 2011. Retrospective cohort study of the risk of obesity among shift workers: findings from the Industry-based Shift Workers’ Health study, Japan. Occup Environ Med, 68, 327-31.

KUBO, T., OZASA, K., MIKAMI, K., WAKAI, K., FUJINO, Y., WATANABE, Y., MIKI, T., NAKAO, M., HAYASHI, K., SUZUKI, K., MORI, M., WASHIO, M., SAKAUCHI, F., ITO, Y., YOSHIMURA, T. & TAMAKOSHI, A. 2006. Prospective cohort study of the risk of prostate cancer among rotating-shift workers: findings from the Japan collaborative cohort study. Am J Epidemiol, 164, 549-55.

KUSANAGI, H., HIDA, A., SATOH, K., ECHIZENYA, M., SHIMIZU, T., PENDERGAST, J. S., YAMAZAKI, S. & MISHIMA, K. 2008. Expression profiles of 10 circadian clock genes in human peripheral blood mononuclear cells. Neurosci Res, 61, 136-42.

LACK, L., BAILEY, M., LOVATO, N. & WRIGHT, H. 2009. Chronotype differences in circadian rhythms of temperature, melatonin, and sleepiness as measured in a modified constant routine protocol. Nat Sci Sleep, 1, 1-8.

LAPIN, V. 1978. Effects of reserpine on the incidence of 9,10-dimethyl-1,2-benzanthracene-induced tumors in pinealectomised and thymectomised rats. Oncology, 35, 132-5.


148

Chapter 8

LAUSS, M., KRIEGNER, A., VIERLINGER, K., VISNE, I., YILDIZ, A., DILAVEROGLU, E. & NOEHAMMER, C. 2008. Consensus genes of the literature to predict breast cancer recurrence. Breast Cancer Res Treat, 110, 235-44.

LEE, C., ETCHEGARAY, J. P., CAGAMPANG, F. R., LOUDON, A. S. & REPPERT, S. M. 2001. Posttranslational mechanisms regulate the mammalian circadian clock. Cell, 107, 855-67.

LENGACHER, S., NEHIRI-SITAYEB, T., STEINER, N., CARNEIRO, L., FAVROD, C., PREITNER, F., THORENS, B., STEHLE, J. C., DIX, L., PRALONG, F., MAGISTRETTI, P. J. & PELLERIN, L. 2013. Resistance to diet-induced obesity and associated metabolic perturbations in haploinsufficient monocarboxylate transporter 1 mice. PLoS One, 8, e82505.

LENNERNAS, M., HAMBRAEUS, L. & AKERSTEDT, T. 1995. Shift related dietary intake in day and shift workers. Appetite, 25, 253-65.

LI, S. X., LIU, L. J., XU, L. Z., GAO, L., WANG, X. F., ZHANG, J. T. & LU, L. 2013. Diurnal alterations in circadian genes and peptides in major depressive disorder before and after escitalopram treatment. Psychoneuroendocrinology, 38, 2789-99.

LIBERZON, A., SUBRAMANIAN, A., PINCHBACK, R., THORVALDSDOTTIR, H., TAMAYO, P. & MESIROV, J. P. 2011. Molecular signatures database (MSigDB) 3.0. Bioinformatics, 27, 1739-40.

LIIRA, J., VERBEEK, J. H., COSTA, G., DRISCOLL, T. R., SALLINEN, M., ISOTALO, L. K. & RUOTSALAINEN, J. H. 2014. Pharmacological interventions for sleepiness and sleep disturbances caused by shift work. Cochrane Database Syst Rev, 8, CD009776.

LOWREY, P. L. & TAKAHASHI, J. S. 2011. Genetics of circadian rhythms in Mammalian model organisms. Adv Genet, 74, 175-230.

LUDWIG, J. A. & WEINSTEIN, J. N. 2005. Biomarkers in cancer staging, prognosis and treatment selection. Nat Rev Cancer, 5, 845-56.

MAZZOCCOLI, G., MUSCARELLA, L. A., FAZIO, V. M., PIEPOLI, A., PAZIENZA, V., DAGOSTINO, M. P., GIULIANI, F., POLYAKOVA, V. O. & KVETNOY, I. 2011. Antiphase signalling in the neuroendocrine-immune system in healthy humans. Biomed Pharmacother, 65, 275-9.

MCGOWAN, N. M. & COOGAN, A. N. 2013. Circadian and behavioural responses to shift work-like schedules of light/dark in the mouse. J Mol Psychiatry, 1.

MEGDAL, S. P., KROENKE, C. H., LADEN, F., PUKKALA, E. & SCHERNHAMMER, E. S. 2005. Night work and breast cancer risk: a systematic review and meta-analysis. Eur J Cancer, 41, 2023-32.

MESTER, B., BEHRENS, T., DREGER, S., HENSE, S. & FRITSCHI, L. 2010. Occupational causes of testicular cancer in adults. Int J Occup Environ Med, 1, 160-70.

MHATRE, M. C., SHAH, P. N. & JUNEJA, H. S. 1984. Effect of varying photoperiods on mammary morphology, DNA synthesis, and hormone profile in female rats. J Natl Cancer Inst, 72, 1411-6.

MIIDA, T., NAKAMURA, Y., MEZAKI, T., HANYU, O., MARUYAMA, S., HORIKAWA, Y., IZAWA, S., YAMADA, Y., MATSUI, H. & OKADA, M. 2002. LDL-cholesterol and HDL-cholesterol concentrations decrease during the day. Ann Clin Biochem, 39, 241-9.

MILLER, B. H., MCDEARMON, E. L., PANDA, S., HAYES, K. R., ZHANG, J., ANDREWS, J. L., ANTOCH, M. P., WALKER, J. R., ESSER, K. A., HOGENESCH, J. B. & TAKAHASHI, J. S. 2007. Circadian and CLOCK-controlled regulation of the mouse transcriptome and cell proliferation. Proc Natl Acad Sci U S A, 104, 3342-7.

MINAMI, Y., KASUKAWA, T., KAKAZU, Y., IIGO, M., SUGIMOTO, M., IKEDA, S., YASUI, A., VAN DER HORST, G. T., SOGA, T. & UEDA, H. R. 2009. Measurement of internal body time by blood metabolomics. Proc Natl Acad Sci U S A, 106, 9890-5.

MIRICK, D. K. & DAVIS, S. 2008. Melatonin as a biomarker of circadian dysregulation. Cancer Epidemiol Biomarkers Prev, 17, 3306-13.

MIYATAKE, A., MORIMOTO, Y., OISHI, T., HANASAKI, N., SUGITA, Y., IIJIMA, S., TESHIMA, Y., HISHIKAWA, Y. & YAMAMURA, Y. 1980. Circadian rhythm of serum testosterone and its relation to sleep: comparison with the variation in serum luteinizing hormone, prolactin, and cortisol in normal men. J Clin Endocrinol Metab, 51, 1365-71.

MOHAWK, J. A., GREEN, C. B. & TAKAHASHI, J. S. 2012. Central and peripheral circadian clocks in mammals. Annu Rev Neurosci, 35, 445-62.

MOLINA, R., AUGE, J. M., ESCUDERO, J. M., FILELLA, X., ZANON, G., PAHISA, J., FARRUS, B., MUNOZ, M. & VELASCO, M. 2010. Evaluation of tumor markers (HER-2/neu oncoprotein, CEA, and CA 15.3) in patients with locoregional breast cancer: prognostic value. Tumour Biol, 31, 171-80.


149

8

References

MOLLER-LEVET, C. S., ARCHER, S. N., BUCCA, G., LAING, E. E., SLAK, A., KABILJO, R., LO, J. C., SANTHI, N., VON SCHANTZ, M., SMITH, C. P. & DIJK, D. J. 2013. Effects of insufficient sleep on circadian rhythmicity and expression amplitude of the human blood transcriptome. Proc Natl Acad Sci U S A, 110, E1132-41.

MOORE, R. Y. 1983. Organization and function of a central nervous system circadian oscillator: the suprachiasmatic hypothalamic nucleus. Fed Proc, 42, 2783-9.

MOORE, R. Y. & EICHLER, V. B. 1972. Loss of a circadian adrenal corticosterone rhythm following suprachiasmatic lesions in the rat. Brain Res, 42, 201-6.

MOOTHA, V. K., LINDGREN, C. M., ERIKSSON, K. F., SUBRAMANIAN, A., SIHAG, S., LEHAR, J., PUIGSERVER, P., CARLSSON, E., RIDDERSTRALE, M., LAURILA, E., HOUSTIS, N., DALY, M. J., PATTERSON, N., MESIROV, J. P., GOLUB, T. R., TAMAYO, P., SPIEGELMAN, B., LANDER, E. S., HIRSCHHORN, J. N., ALTSHULER, D. & GROOP, L. C. 2003. PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet, 34, 267-73.

MORRIS, C. J., YANG, J. N., GARCIA, J. I., MYERS, S., BOZZI, I., WANG, W., BUXTON, O. M., SHEA, S. A. & SCHEER, F. A. 2015. Endogenous circadian system and circadian misalignment impact glucose tolerance via separate mechanisms in humans. Proc Natl Acad Sci U S A, 112, E2225-34.

MORTOLA, J. F., LAUGHLIN, G. A. & YEN, S. S. 1992. A circadian rhythm of serum follicle-stimulating hormone in women. J Clin Endocrinol Metab, 75, 861-4.

MURAKAMI, Y., KOHYAMA, N., KOBAYASHI, Y., OHBAYASHI, M., OHTANI, H., SAWADA, Y. & YAMAMOTO, T. 2005. Functional characterization of human monocarboxylate transporter 6 (SLC16A5). Drug Metab Dispos, 33, 1845-51.

MYLLYKANGAS, S., HIMBERG, J., BOHLING, T., NAGY, B., HOLLMEN, J. & KNUUTILA, S. 2006. DNA copy number amplification profiling of human neoplasms. Oncogene, 25, 7324-32.

NABE-NIELSEN, K., QUIST, H. G., GARDE, A. H. & AUST, B. 2011. Shiftwork and changes in health behaviors. J Occup Environ Med, 53, 1413-7.

NANDI, S., GUZMAN, R. C. & YANG, J. 1995. Hormones and mammary carcinogenesis in mice, rats, and humans: a unifying hypothesis. Proc Natl Acad Sci U S A, 92, 3650-7.

NANJAPPA, V., THOMAS, J. K., MARIMUTHU, A., MUTHUSAMY, B., RADHAKRISHNAN, A., SHARMA, R., AHMAD KHAN, A., BALAKRISHNAN, L., SAHASRABUDDHE, N. A., KUMAR, S., JHAVERI, B. N., SHETH, K. V., KUMAR KHATANA, R., SHAW, P. G., SRIKANTH, S. M., MATHUR, P. P., SHANKAR, S., NAGARAJA, D., CHRISTOPHER, R., MATHIVANAN, S., RAJU, R., SIRDESHMUKH, R., CHATTERJEE, A., SIMPSON, R. J., HARSHA, H. C., PANDEY, A. & PRASAD, T. S. 2014. Plasma Proteome Database as a resource for proteomics research: 2014 update. Nucleic Acids Res, 42, D959-65.

NEIL, S. E., PAHWA, M., DEMERS, P. A. & GOTAY, C. C. 2014. Health-related interventions among night shift workers: a critical review of the literature. Scand J Work Environ Health.

NICOLAU, G. Y. & HAUS, E. 1989. Chronobiology of the endocrine system. Endocrinologie, 27, 153-83.

NIKLOWITZ, P., ANDLER, W. & MENKE, T. 2006. Coenzyme Q10 concentration in plasma and blood cells: what about diurnal changes? Biofactors, 28, 47-54.

NIU, S. F., CHUNG, M. H., CHU, H., TSAI, J. C., LIN, C. C., LIAO, Y. M., OU, K. L., O’BRIEN, A. P. & CHOU, K. R. 2015. Differences in cortisol profiles and circadian adjustment time between nurses working night shifts and regular day shifts: A prospective longitudinal study. Int J Nurs Stud, 52, 1193-201.

NORBERG, T., KLAAR, S., KARF, G., NORDGREN, H., HOLMBERG, L. & BERGH, J. 2001. Increased p53 mutation frequency during tumor progression--results from a breast cancer cohort. Cancer Res, 61, 8317-21.

NORUM, L. F., ERIKSTEIN, B. & NUSTAD, K. 2001. Elevated CA125 in breast cancer--A sign of advanced disease. Tumour Biol, 22, 223-8.

OHTA, H., YAMAZAKI, S. & MCMAHON, D. G. 2005. Constant light desynchronizes mammalian clock neurons. Nat Neurosci, 8, 267-9.

OIKE, H., SAKURAI, M., IPPOUSHI, K. & KOBORI, M. 2015. Time-fixed feeding prevents obesity induced by chronic advances of light/dark cycles in mouse models of jet-lag/shift work. Biochem Biophys Res Commun, 465, 556-61.

OISHI, K. 2009. Disrupted light-dark cycle induces obesity with hyperglycemia in genetically intact animals. Neuro Endocrinol Lett, 30, 458-61.

OLLILA, H. M., UTGE, S., KRONHOLM, E., AHO, V., VAN LEEUWEN, W., SILANDER, K., PARTONEN, T., PEROLA, M., KAPRIO, J., SALOMAA, V., SALLINEN, M., HARMA, M., PORKKA-HEISKANEN, T. & PAUNIO, T. 2012. TRIB1 constitutes a molecular link between regulation of sleep and lipid metabolism in humans. Transl Psychiatry, 2, e97.


150

Chapter 8

OPPERHUIZEN, A. L., VAN KERKHOF, L. W., PROPER, K. I., RODENBURG, W. & KALSBEEK, A. 2015. Rodent models to study the metabolic effects of shiftwork in humans. Front Pharmacol, 6, 50.

OPSTAL-VAN WINDEN, A. W., KROP, E. J., KAREDAL, M. H., GAST, M. C., LINDH, C. H., JEPPSSON, M. C., JONSSON, B. A., GROBBEE, D. E., PEETERS, P. H., BEIJNEN, J. H., VAN GILS, C. H. & VERMEULEN, R. C. 2011. Searching for early breast cancer biomarkers by serum protein profiling of pre-diagnostic serum; a nested case-control study. BMC Cancer, 11, 381.

ORTH-GOMER, K. 1982. Intervention on coronary risk factors by changing working conditions of Swedish policemen. Act Nerv Super (Praha), Suppl 3, 223-9.

OSTROWSKA, Z., ZWIRSKA-KORCZALA, K., PARDELA, M., DROZDZ, M., KOS-KUDLA, B. & BUNTNER, B. 1998. Circadian variations of androstenedione, dehydroepiandrosterone sulfate and free testosterone in obese women with menstrual disturbances. Endocr Regul, 32, 169-176.

PACK, A. I., GALANTE, R. J., MAISLIN, G., CATER, J., METAXAS, D., LU, S., ZHANG, L., VON SMITH, R., KAY, T., LIAN, J., SVENSON, K. & PETERS, L. L. 2007. Novel method for high-throughput phenotyping of sleep in mice. Physiol Genomics, 28, 232-8.

PAN, A., SCHERNHAMMER, E. S., SUN, Q. & HU, F. B. 2011. Rotating night shift work and risk of type 2 diabetes: two prospective cohort studies in women. PLoS Med, 8, e1001141.

PANDA, S., ANTOCH, M. P., MILLER, B. H., SU, A. I., SCHOOK, A. B., STRAUME, M., SCHULTZ, P. G., KAY, S. A., TAKAHASHI, J. S. & HOGENESCH, J. B. 2002. Coordinated transcription of key pathways in the mouse by the circadian clock. Cell, 109, 307-20.

PAPANTONIOU, K., POZO, O. J., ESPINOSA, A., MARCOS, J., CASTANO-VINYALS, G., BASAGANA, X., RIBAS, F. C., MIRABENT, J., MARTIN, J., CARENYS, G., MARTIN, C. R., MIDDLETON, B., SKENE, D. J. & KOGEVINAS, M. 2014. Circadian variation of melatonin, light exposure, and diurnal preference in day and night shift workers of both sexes. Cancer Epidemiol Biomarkers Prev, 23, 1176-86.

PARSONS, M. J., MOFFITT, T. E., GREGORY, A. M., GOLDMAN-MELLOR, S., NOLAN, P. M., POULTON, R. & CASPI, A. 2014. Social jetlag, obesity and metabolic disorder: investigation in a cohort study. Int J Obes (Lond).

PENNINGS, J. L., RODENBURG, W., IMHOLZ, S., KOSTER, M. P., VAN OOSTROM, C. T., BREIT, T. M., SCHIELEN, P. C. & DE VRIES, A. 2011. Gene expression profiling in a mouse model identifies fetal liver- and placenta-derived potential biomarkers for Down Syndrome screening. PLoS One, 6, e18866.

PENNINGS, J. L., SILJEE, J. E., IMHOLZ, S., KUC, S., DE VRIES, A., SCHIELEN, P. C. & RODENBURG, W. 2014. Comparison of different blood collection, sample matrix, and immunoassay methods in a prenatal screening setting. Dis Markers, 2014, 509821.

PETROV, M. E., CLARK, C. B., MOLZOF, H. E., JOHNSON, R. L., JR., CROPSEY, K. L. & GAMBLE, K. L. 2014. Sleep Strategies of Night-Shift Nurses on Days Off: Which Ones are Most Adaptive? Front Neurol, 5, 277.

PFEFFER, M., WICHT, H., VON GALL, C. & KORF, H. W. 2015. Owls and larks in mice. Front Neurol, 6, 101.

PHILLIPS, D. J., SAVENKOVA, M. I. & KARATSOREOS, I. N. 2015. Environmental disruption of the circadian clock leads to altered sleep and immune responses in mouse. Brain Behav Immun, 47, 14-23.

PICKERING, T. G. 1990. The clinical significance of diurnal blood pressure variations. Dippers and nondippers. Circulation, 81, 700-2.

PIJPE, A., SLOTTJE, P., VAN PELT, C., STEHMANN, F., KROMHOUT, H., VAN LEEUWEN, F. E., VERMEULEN, R. C. & ROOKUS, M. A. 2014. The Nightingale study: rationale, study design and baseline characteristics of a prospective cohort study on shift work and breast cancer risk among nurses. BMC Cancer, 14, 47.

PITTERI, S. J., FACA, V. M., KELLY-SPRATT, K. S., KASARDA, A. E., WANG, H., ZHANG, Q., NEWCOMB, L., KRASNOSELSKY, A., PACZESNY, S., CHOI, G., FITZGIBBON, M., MCINTOSH, M. W., KEMP, C. J. & HANASH, S. M. 2008. Plasma proteome profiling of a mouse model of breast cancer identifies a set of up-regulated proteins in common with human breast cancer cells. J Proteome Res, 7, 1481-9.

PLYMATE, S. R., TENOVER, J. S. & BREMNER, W. J. 1989. Circadian variation in testosterone, sex hormone-binding globulin, and calculated non-sex hormone-binding globulin bound testosterone in healthy young and elderly men. J Androl, 10, 366-71.

PONTEN, F., JIRSTROM, K. & UHLEN, M. 2008. The Human Protein Atlas--a tool for pathology. J Pathol, 216, 387-93.

POOLA, I., DEWITTY, R. L., MARSHALLECK, J. J., BHATNAGAR, R., ABRAHAM, J. & LEFFALL, L. D. 2005. Identification of MMP-1 as a putative breast cancer predictive marker by global gene expression analysis. Nat Med, 11, 481-3.

PUKKALA, E., AUVINEN, A. & WAHLBERG, G. 1995. Incidence of cancer among Finnish airline cabin attendants, 1967-92. BMJ, 311, 649-52.


151

8

References

REDMAN, J. R. 1997. Circadian entrainment and phase shifting in mammals with melatonin. J Biol Rhythms, 12, 581-7.

REFINETTI, R. & MENAKER, M. 1992. The circadian rhythm of body temperature. Physiol Behav, 51, 613-37.

REPPERT, S. M. & WEAVER, D. R. 2002. Coordination of circadian timing in mammals. Nature, 418, 935-41.

REVILLION, F., BONNETERRE, J. & PEYRAT, J. P. 1998. ERBB2 oncogene in human breast cancer and its clinical significance. Eur J Cancer, 34, 791-808.

REYNOLDS, P., CONE, J., LAYEFSKY, M., GOLDBERG, D. E. & HURLEY, S. 2002. Cancer incidence in California flight attendants (United States). Cancer Causes Control, 13, 317-24.

RICHTER, K. D., ACKER, J., SCHOLZ, F. & NIKLEWSKI, G. 2010. Health promotion and work: prevention of shift work disorders in companies. EPMA J, 1, 611-8.

RIEDEL, M., HOEFT, B., BLUM, W. F., VON ZUR MUHLEN, A. & BRABANT, G. 1995. Pulsatile growth hormone secretion in normal-weight and obese men: differential metabolic regulation during energy restriction. Metabolism, 44, 605-10.

RIFKIN, R., MUKHERJEE, S., TAMAYO, P., RAMASWAMY, S., YEANG, S., ANGELO, M., REICH, M., POGGIO, T., LANDER, E. S., GOLUB, T. R. & MESIROV, J. P. 2003. An analytical method for multi-class molecular cancer classification. Siam Review, 45, 706-723.

ROENNEBERG, T., ALLEBRANDT, K. V., MERROW, M. & VETTER, C. 2012. Social jetlag and obesity. Curr Biol, 22, 939-43.

ROENNEBERG, T., KELLER, L. K., FISCHER, D., MATERA, J. L., VETTER, C. & WINNEBECK, E. C. 2015. Human activity and rest in situ. Methods Enzymol, 552, 257-83.

RUGGIERO, J. S. & REDEKER, N. S. 2014. Effects of napping on sleepiness and sleep-related performance deficits in night-shift workers: a systematic review. Biol Res Nurs, 16, 134-42.

SACK, R. L., BLOOD, M. L. & LEWY, A. J. 1992. Melatonin rhythms in night shift workers. Sleep, 15, 434-41.

SAGE, D., GANEM, J., GUILLAUMOND, F., LAFORGE-ANGLADE, G., FRANCOIS-BELLAN, A. M., BOSLER, O. & BECQUET, D. 2004. Influence of the corticosterone rhythm on photic entrainment of locomotor activity in rats. J Biol Rhythms, 19, 144-56.

SALGADO-DELGADO, R., ANGELES-CASTELLANOS, M., BUIJS, M. R. & ESCOBAR, C. 2008. Internal desynchronization in a model of night-work by forced activity in rats. Neuroscience, 154, 922-31.

SALGADO-DELGADO, R., ANGELES-CASTELLANOS, M., SADERI, N., BUIJS, R. M. & ESCOBAR, C. 2010. Food intake during the normal activity phase prevents obesity and circadian desynchrony in a rat model of night work. Endocrinology, 151, 1019-29.

SANCAR, A., LINDSEY-BOLTZ, L. A., KANG, T. H., REARDON, J. T., LEE, J. H. & OZTURK, N. 2010. Circadian clock control of the cellular response to DNA damage. FEBS Lett, 584, 2618-25.

SASSEVILLE, A. & HEBERT, M. 2010. Using blue-green light at night and blue-blockers during the day to improves adaptation to night work: a pilot study. Prog Neuropsychopharmacol Biol Psychiatry, 34, 1236-42.

SCHERNHAMMER, E. S., KROENKE, C. H., LADEN, F. & HANKINSON, S. E. 2006. Night work and risk of breast cancer. Epidemiology, 17, 108-11.

SCHERNHAMMER, E. S., LADEN, F., SPEIZER, F. E., WILLETT, W. C., HUNTER, D. J., KAWACHI, I., FUCHS, C. S. & COLDITZ, G. A. 2003. Night-shift work and risk of colorectal cancer in the nurses’ health study. J Natl Cancer Inst, 95, 825-8.

SCHOENHERR, R. M., KELLY-SPRATT, K. S., LIN, C., WHITEAKER, J. R., LIU, T., HOLZMAN, T., COLEMAN, I., FENG, L. C., LORENTZEN, T. D., KRASNOSELSKY, A. L., WANG, P., LIU, Y., GURLEY, K. E., AMON, L. M., SCHEPMOES, A. A., MOORE, R. J., CAMP, D. G., 2ND, CHODOSH, L. A., SMITH, R. D., NELSON, P. S., MCINTOSH, M. W., KEMP, C. J. & PAULOVICH, A. G. 2011. Proteome and transcriptome profiles of a Her2/Neu-driven mouse model of breast cancer. Proteomics Clin Appl, 5, 179-88.

SCHUETZ, C. S., BONIN, M., CLARE, S. E., NIESELT, K., SOTLAR, K., WALTER, M., FEHM, T., SOLOMAYER, E., RIESS, O., WALLWIENER, D., KUREK, R. & NEUBAUER, H. J. 2006. Progression-specific genes identified by expression profiling of matched ductal carcinomas in situ and invasive breast tumors, combining laser capture microdissection and oligonucleotide microarray analysis. Cancer Res, 66, 5278-86.

SCHWARTZ, W. J. & ZIMMERMAN, P. 1990. Circadian timekeeping in BALB/c and C57BL/6 inbred mouse strains. J Neurosci, 10, 3685-94.

SEGAL, E., FRIEDMAN, N., KOLLER, D. & REGEV, A. 2004. A module map showing conditional activity of expression modules in cancer. Nat.Genet., 36, 1090-1098.


152

Chapter 8

SHAH, P. N., MHATRE, M. C. & KOTHARI, L. S. 1984. Effect of melatonin on mammary carcinogenesis in intact and pinealectomized rats in varying photoperiods. Cancer Res, 44, 3403-7.

SHAW, P. J., CIRELLI, C., GREENSPAN, R. J. & TONONI, G. 2000. Correlates of sleep and waking in Drosophila melanogaster. Science, 287, 1834-7.

SORLIE, T., PEROU, C. M., TIBSHIRANI, R., AAS, T., GEISLER, S., JOHNSEN, H., HASTIE, T., EISEN, M. B., VAN DE RIJN, M., JEFFREY, S. S., THORSEN, T., QUIST, H., MATESE, J. C., BROWN, P. O., BOTSTEIN, D., EYSTEIN LONNING, P. & BORRESEN-DALE, A. L. 2001. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A, 98, 10869-74.

SOTIRIOU, C. & PUSZTAI, L. 2009. Gene-expression signatures in breast cancer. N Engl J Med, 360, 790-800.

SOTIRIOU, C., WIRAPATI, P., LOI, S., HARRIS, A., FOX, S., SMEDS, J., NORDGREN, H., FARMER, P., PRAZ, V., HAIBE-KAINS, B., DESMEDT, C., LARSIMONT, D., CARDOSO, F., PETERSE, H., NUYTEN, D., BUYSE, M., VAN DE VIJVER, M. J., BERGH, J., PICCART, M. & DELORENZI, M. 2006. Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis. J Natl Cancer Inst, 98, 262-72.

SPRATT, D. I., O’DEA, L. S., SCHOENFELD, D., BUTLER, J., RAO, P. N. & CROWLEY, W. F., JR. 1988. Neuroendocrine-gonadal axis in men: frequent sampling of LH, FSH, and testosterone. Am J Physiol, 254, E658-66.

STANDISH, L. J., SWEET, E. S., NOVACK, J., WENNER, C. A., BRIDGE, C., NELSON, A., MARTZEN, M. & TORKELSON, C. 2008. Breast cancer and the immune system. J Soc Integr Oncol, 6, 158-68.

STEVENS, R. G. 2005. Circadian disruption and breast cancer: from melatonin to clock genes. Epidemiology, 16, 254-8.

STORCH, K. F., LIPAN, O., LEYKIN, I., VISWANATHAN, N., DAVIS, F. C., WONG, W. H. & WEITZ, C. J. 2002. Extensive and divergent circadian gene expression in liver and heart. Nature, 417, 78-83.

STURGEON, C. M., DUFFY, M. J., STENMAN, U. H., LILJA, H., BRUNNER, N., CHAN, D. W., BABAIAN, R., BAST, R. C., JR., DOWELL, B., ESTEVA, F. J., HAGLUND, C., HARBECK, N., HAYES, D. F., HOLTEN-ANDERSEN, M., KLEE, G. G., LAMERZ, R., LOOIJENGA, L. H., MOLINA, R., NIELSEN, H. J., RITTENHOUSE, H., SEMJONOW, A., SHIH IE, M., SIBLEY, P., SOLETORMOS, G., STEPHAN, C., SOKOLL, L., HOFFMAN, B. R. & DIAMANDIS, E. P. 2008. National Academy of Clinical Biochemistry laboratory medicine practice guidelines for use of tumor markers in testicular, prostate, colorectal, breast, and ovarian cancers. Clin Chem, 54, e11-79.

SUBRAMANIAN, A. & KOTHARI, L. 1991. Suppressive effect by melatonin on different phases of 9,10-dimethyl-1,2-benzanthracene (DMBA)-induced rat mammary gland carcinogenesis. Anticancer Drugs, 2, 297-303.

SUBRAMANIAN, A., TAMAYO, P., MOOTHA, V. K., MUKHERJEE, S., EBERT, B. L., GILLETTE, M. A., PAULOVICH, A., POMEROY, S. L., GOLUB, T. R., LANDER, E. S. & MESIROV, J. P. 2005. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A, 102, 15545-50.

SUJINO, M., FURUKAWA, K., KOINUMA, S., FUJIOKA, A., NAGANO, M., IIGO, M. & SHIGEYOSHI, Y. 2012. Differential entrainment of peripheral clocks in the rat by glucocorticoid and feeding. Endocrinology, 153, 2277-86.

SUWAZONO, Y., SAKATA, K., OKUBO, Y., HARADA, H., OISHI, M., KOBAYASHI, E., UETANI, M., KIDO, T. & NOGAWA, K. 2006. Long-term longitudinal study on the relationship between alternating shift work and the onset of diabetes mellitus in male Japanese workers. J Occup Environ Med, 48, 455-61.

TAKIMOTO, M., HAMADA, A., TOMODA, A., OHDO, S., OHMURA, T., SAKATO, H., KAWATANI, J., JODOI, T., NAKAGAWA, H., TERAZONO, H., KOYANAGI, S., HIGUCHI, S., KIMURA, M., TUKIKAWA, H., IRIE, S., SAITO, H. & MIIKE, T. 2005. Daily expression of clock genes in whole blood cells in healthy subjects and a patient with circadian rhythm sleep disorder. Am J Physiol Regul Integr Comp Physiol, 289, R1273-9.

TAMARKIN, L., COHEN, M., ROSELLE, D., REICHERT, C., LIPPMAN, M. & CHABNER, B. 1981. Melatonin inhibition and pinealectomy enhancement of 7,12-dimethylbenz(a)anthracene-induced mammary tumors in the rat. Cancer Res, 41, 4432-6.

TAMPELLINI, M., BERRUTI, A., BITOSSI, R., GORZEGNO, G., ALABISO, I., BOTTINI, A., FARRIS, A., DONADIO, M., SAROBBA, M. G., MANZIN, E., DURANDO, A., DEFABIANI, E., DE MATTEIS, A., ARDINE, M., CASTIGLIONE, F., DANESE, S., BERTONE, E., ALABISO, O., MASSOBRIO, M. & DOGLIOTTI, L. 2006. Prognostic significance of changes in CA 15-3 serum levels during chemotherapy in metastatic breast cancer patients. Breast Cancer Res Treat, 98, 241-8.

TEBOUL, M., BARRAT-PETIT, M. A., LI, X. M., CLAUSTRAT, B., FORMENTO, J. L., DELAUNAY, F., LEVI, F. & MILANO, G. 2005. Atypical patterns of circadian clock gene expression in human peripheral blood mononuclear cells. J Mol Med (Berl), 83, 693-9.

TIBSHIRANI, R., HASTIE, T., NARASIMHAN, B. & CHU, G. 2002. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci U S A, 99, 6567-72.


153

8

References

TRAVLOS, G. S., WILSON, R. E., MURRELL, J. A., CHIGNELL, C. F. & BOORMAN, G. A. 2001. The effect of short intermittent light exposures on the melatonin circadian rhythm and NMU-induced breast cancer in female F344/N rats. Toxicol Pathol, 29, 126-36.

TUCKER, P., SMITH, L., MACDONALD, I. & FOLKARD, S. 2000. Effects of direction of rotation in continuous and discontinuous 8 hour shift systems. Occup Environ Med, 57, 678-84.

UEDA, H. R., CHEN, W., MINAMI, Y., HONMA, S., HONMA, K., IINO, M. & HASHIMOTO, S. 2004. Molecular-timetable methods for detection of body time and rhythm disorders from single-time-point genome-wide expression profiles. Proc Natl Acad Sci U S A, 101, 11227-32.

UPTON, A. C., ALBERT, R. E. & BURNS, F. J. 1986. Radiation Carcinogenesis, New York, Elsevier.

URAY, I. P., LIANG, Y. & HYDER, S. M. 2004. Estradiol down-regulates CD36 expression in human breast cancer cells. Cancer Lett, 207, 101-7.

VAN DEN POL, A. N. 1980. The hypothalamic suprachiasmatic nucleus of rat: intrinsic anatomy. J Comp Neurol, 191, 661-702.

VAN DER RHEE, H., COEBERGH, J. W. & DE VRIES, E. 2013. Is prevention of cancer by sun exposure more than just the effect of vitamin D? A systematic review of epidemiological studies. Eur J Cancer, 49, 1422-36.

VAN DER VEEN, J. W., PRONK, T. E., VAN LOVEREN, H. & EZENDAM, J. 2013. Applicability of a keratinocyte gene signature to predict skin sensitizing potential. Toxicol In Vitro, 27, 314-22.

VAN DYCKE, K. C., RODENBURG, W., VAN OOSTROM, C. T., VAN KERKHOF, L. W., PENNINGS, J. L., ROENNEBERG, T., VAN STEEG, H. & VAN DER HORST, G. T. 2015. Chronically Alternating Light Cycles Increase Breast Cancer Risk in Mice. Curr Biol, 25, 1932-7.

VANDEBRIEL, R. J., PENNINGS, J. L., BAKEN, K. A., PRONK, T. E., BOORSMA, A., GOTTSCHALK, R. & VAN LOVEREN, H. 2010. Keratinocyte gene expression profiles discriminate sensitizing and irritating compounds. Toxicol Sci, 117, 81-9.

VANHECKE, E., ADRIAENSSENS, E., VERBEKE, S., MEIGNAN, S., GERMAIN, E., BERTEAUX, N., NURCOMBE, V., LE BOURHIS, X. & HONDERMARCK, H. 2011. Brain-derived neurotrophic factor and neurotrophin-4/5 are expressed in breast cancer and can be targeted to inhibit tumor cell survival. Clin Cancer Res, 17, 1741-52.

VARLEY, J. M., EVANS, D. G. & BIRCH, J. M. 1997. Li-Fraumeni syndrome--a molecular and clinical review. Br J Cancer, 76, 1-14.

VELDHUIS, J. D., CHRISTIANSEN, E., EVANS, W. S., KOLP, L. A., ROGOL, A. D. & JOHNSON, M. L. 1988. Physiological profiles of episodic progesterone release during the midluteal phase of the human menstrual cycle: analysis of circadian and ultradian rhythms, discrete pulse properties, and correlations with simultaneous luteinizing hormone release. J Clin Endocrinol Metab, 66, 414-21.

VERSCHUREN, W. M., BLOKSTRA, A., PICAVET, H. S. & SMIT, H. A. 2008. Cohort profile: the Doetinchem Cohort Study. Int J Epidemiol, 37, 1236-41.

VETTER, C., FISCHER, D., MATERA, J. L. & ROENNEBERG, T. 2015. Aligning work and circadian time in shift workers improves sleep and reduces circadian disruption. Curr Biol, 25, 1-5.

VIITASALO, K., KUOSMA, E., LAITINEN, J. & HARMA, M. 2008. Effects of shift rotation and the flexibility of a shift system on daytime alertness and cardiovascular risk factors. Scand J Work Environ Health, 34, 198-205.

VINOGRADOVA, I. A., ANISIMOV, V. N., BUKALEV, A. V., ILYUKHA, V. A., KHIZHKIN, E. A., LOTOSH, T. A., SEMENCHENKO, A. V. & ZABEZHINSKI, M. A. 2010. Circadian disruption induced by light-at-night accelerates aging and promotes tumorigenesis in young but not in old rats. Aging (Albany NY), 2, 82-92.

VINOGRADOVA, I. A., ANISIMOV, V. N., BUKALEV, A. V., SEMENCHENKO, A. V. & ZABEZHINSKI, M. A. 2009. Circadian disruption induced by light-at-night accelerates aging and promotes tumorigenesis in rats. Aging (Albany NY), 1, 855-65.

WAGNER, K. U., WALL, R. J., ST-ONGE, L., GRUSS, P., WYNSHAW-BORIS, A., GARRETT, L., LI, M., FURTH, P. A. & HENNIGHAUSEN, L. 1997. Cre-mediated gene deletion in the mammary gland. Nucleic Acids Res, 25, 4323-30.

WANG, F., YEUNG, K. L., CHAN, W. C., KWOK, C. C., LEUNG, S. L., WU, C., CHAN, E. Y., YU, I. T., YANG, X. R. & TSE, L. A. 2013. A meta-analysis on dose-response relationship between night shift work and the risk of breast cancer. Ann Oncol.

WANG, F., ZHANG, L., ZHANG, Y., ZHANG, B., HE, Y., XIE, S., LI, M., MIAO, X., CHAN, E. Y., TANG, J. L., WONG, M. C., LI, Z., YU, I. T. & TSE, L. A. 2014. Meta-analysis on night shift work and risk of metabolic syndrome. Obes Rev, 15, 709-20.


154

Chapter 8

WANG, X. S., TRAVIS, R. C., REEVES, G., GREEN, J., ALLEN, N. E., KEY, T. J., RODDAM, A. W. & BERAL, V. 2012. Characteristics of the Million Women Study participants who have and have not worked at night. Scand J Work Environ Health, 38, 590-9.

WARMAN, V. L., DIJK, D. J., WARMAN, G. R., ARENDT, J. & SKENE, D. J. 2003. Phase advancing human circadian rhythms with short wavelength light. Neurosci Lett, 342, 37-40.

WHITEAKER, J. R., LIN, C., KENNEDY, J., HOU, L., TRUTE, M., SOKAL, I., YAN, P., SCHOENHERR, R. M., ZHAO, L., VOYTOVICH, U. J., KELLY-SPRATT, K. S., KRASNOSELSKY, A., GAFKEN, P. R., HOGAN, J. M., JONES, L. A., WANG, P., AMON, L., CHODOSH, L. A., NELSON, P. S., MCINTOSH, M. W., KEMP, C. J. & PAULOVICH, A. G. 2011. A targeted proteomics-based pipeline for verification of biomarkers in plasma. Nat Biotechnol, 29, 625-34.

WHITEAKER, J. R., ZHANG, H., ZHAO, L., WANG, P., KELLY-SPRATT, K. S., IVEY, R. G., PIENING, B. D., FENG, L. C., KASARDA, E., GURLEY, K. E., ENG, J. K., CHODOSH, L. A., KEMP, C. J., MCINTOSH, M. W. & PAULOVICH, A. G. 2007. Integrated pipeline for mass spectrometry-based discovery and confirmation of biomarkers demonstrated in a mouse model of breast cancer. J Proteome Res, 6, 3962-75.

WHO. 2013. Fact sheet Diabetes [Online]. Available: http://www.who.int/mediacentre/factsheets/fs312/en/.

WIJNHOVEN, S. W., ZWART, E., SPEKSNIJDER, E. N., BEEMS, R. B., OLIVE, K. P., TUVESON, D. A., JONKERS, J., SCHAAP, M. M., VAN DEN BERG, J., JACKS, T., VAN STEEG, H. & DE VRIES, A. 2005. Mice expressing a mammary gland-specific R270H mutation in the p53 tumor suppressor gene mimic human breast cancer development. Cancer Res, 65, 8166-73.

WU, C., OROZCO, C., BOYER, J., LEGLISE, M., GOODALE, J., BATALOV, S., HODGE, C. L., HAASE, J., JANES, J., HUSS, J. W., III & SU, A. I. 2009. BioGPS: an extensible and customizable portal for querying and organizing gene annotation resources. Genome Biol., 10, R130.

WU, J., DAUCHY, R. T., TIRRELL, P. C., WU, S. S., LYNCH, D. T., JITAWATANARAT, P., BURRINGTON, C. M., DAUCHY, E. M., BLASK, D. E. & GREENE, M. W. 2011. Light at night activates IGF-1R/PDK1 signaling and accelerates tumor growth in human breast cancer xenografts. Cancer Res, 71, 2622-31.

WU, P., BLAIR, P. V., SATO, J., JASKIEWICZ, J., POPOV, K. M. & HARRIS, R. A. 2000. Starvation increases the amount of pyruvate dehydrogenase kinase in several mammalian tissues. Arch Biochem Biophys, 381, 1-7.

XU, B., GOULDING, E. H., ZANG, K., CEPOI, D., CONE, R. D., JONES, K. R., TECOTT, L. H. & REICHARDT, L. F. 2003. Brain-derived neurotrophic factor regulates energy balance downstream of melanocortin-4 receptor. Nat Neurosci, 6, 736-42.

YANG, X., WOOD, P. A., ANSELL, C. M., QUITON, D. F., OH, E. Y., DU-QUITON, J. & HRUSHESKY, W. J. 2009a. The circadian clock gene Per1 suppresses cancer cell proliferation and tumor growth at specific times of day. Chronobiol Int, 26, 1323-39.

YANG, X., WOOD, P. A., OH, E. Y., DU-QUITON, J., ANSELL, C. M. & HRUSHESKY, W. J. 2009b. Down regulation of circadian clock gene Period 2 accelerates breast cancer growth by altering its daily growth rhythm. Breast Cancer Res Treat, 117, 423-31.

YOON, J. A., HAN, D. H., NOH, J. Y., KIM, M. H., SON, G. H., KIM, K., KIM, C. J., PAK, Y. K. & CHO, S. 2012. Meal time shift disturbs circadian rhythmicity along with metabolic and behavioral alterations in mice. PLoS One, 7, e44053.

YOSHINO, J., ALMEDA-VALDES, P., PATTERSON, B. W., OKUNADE, A. L., IMAI, S., MITTENDORFER, B. & KLEIN, S. 2014. Diurnal variation in insulin sensitivity of glucose metabolism is associated with diurnal variations in whole-body and cellular fatty acid metabolism in metabolically normal women. J Clin Endocrinol Metab, 99, E1666-70.

YOU, S., WOOD, P. A., XIONG, Y., KOBAYASHI, M., DU-QUITON, J. & HRUSHESKY, W. J. 2005. Daily coordination of cancer growth and circadian clock gene expression. Breast Cancer Res Treat, 91, 47-60.

ZEIWAR, M. M., ZAKI, S. M., MOHAMMAD, L. A., ZIDAN, A. A. & EL NAGAR, M. R. 2007. HER-2 gene amplification, serum nucleosomes, CEA and CA15.3 tumor markers in breast cancer patients. Egypt J Immunol, 14, 29-41.

ZHANG, R., LAHENS, N. F., BALLANCE, H. I., HUGHES, M. E. & HOGENESCH, J. B. 2014a. A circadian gene expression atlas in mammals: Implications for biology and medicine. Proc Natl Acad Sci U S A.

ZHANG, S., HULVER, M. W., MCMILLAN, R. P., CLINE, M. A. & GILBERT, E. R. 2014b. The pivotal role of pyruvate dehydrogenase kinases in metabolic flexibility. Nutr Metab (Lond), 11, 10.


APPENDIX 1

A day and night difference in the response of the

hepatic transcriptome to cyclophosphamide

treatment

Kirsten C.G. van Dycke*, Romana M. Nijman*, Paul F.K. Wackers, Martijs J. Jonker,

Wendy Rodenburg, Conny T.M. van Oostrom, Daniela C.F. Salvatori, Timo M. Breit,

Harry van Steeg, Mirjam Luijten, and Gijsbertus T.J. van der Horst


Archives of Toxicology, 2014, 89(2):221-31


Appendix 1

158

Abstract

Application of omics-based technologies is a widely used approach in research

aiming to improve testing strategies for human health risk assessment. In most

of these studies, however, temporal variations in gene expression caused by the

circadian clock are a commonly neglected pitfall. In the present study, we investigated

the impact of the circadian clock on the response of the hepatic transcriptome

after exposure of mice to the chemotherapeutic agent cyclophosphamide (CP).

Analysis of the data without considering clock progression revealed common

responses in terms of regulated pathways between light and dark phase exposure,

including DNA damage, oxidative stress, and a general immune response. The

overall response however was stronger in mice exposed during the day. Use of

time-matched controls, thereby eliminating non-CP responsive circadian clock

controlled genes, showed that this difference in response was actually even more

pronounced: CP-related responses were only identified in mice exposed during

the day. Only minor differences were found in acute toxicity pathways, namely

lymphocyte counts and kidney weights, indicating that gene expression is subject

to time of day effects. This study is the first to highlight the impact of the circadian

clock on the identification of toxic responses by omics approaches.


Time of day effect of CP

159

A1

Introduction

Human health risk assessment still requires the use of many laboratory animals

to identify hazard and estimate risk of adverse health effects induced by

environmental and pharmaceutical chemicals. Toxicogenomics approaches are

applied to develop and validate risk profiles of biomarkers, identified by profiling

the changes in transcript, protein and/or metabolite levels in in vivo exposed

animals. The combined use of ‘omics’ technology and bioinformatics tools across

platforms and species further supports the identification of common patterns in

the toxic response, as well as the establishment of mechanism-based predictive

biomarker sets. However, a commonly neglected pitfall of this approach is that

temporal or daily variations in gene expression, as well as protein and metabolite

levels, and as a consequence the time of day of exposure, may affect the outcome

of a toxicogenomics experiment.

To anticipate the day-night cycle, dictated by the Earth’s 24-hour rotation around

its axis, most if not all life forms have developed a circadian clock (from the Latin

circa diem, “about a day”) with a near 24-hour periodicity. This internal timekeeping

system imposes day-night rhythms on behavior, physiology and metabolism: e.g.

sleep-wake cycle, body temperature, blood pressure, energy metabolism and

hormone levels. In this way, organisms adjust specific body functions to their

special physiological needs during the 24-hour day (Lowrey and Takahashi 2011;

Mohawk et al. 2012). In mammals, the circadian system consists of a master clock

in the brain, and peripheral clocks in virtually all other cells and tissues. The master

clock, comprised of the neurons of the suprachiasmatic nuclei (SCN), is sensitive to

retinal light stimuli, allowing the clock to keep phase with the light-dark cycle. In

turn, the SCN synchronize peripheral clocks through hormonal and neural factors

(Balsalobre et al. 1998; Reppert and Weaver 2002; Weaver 1998; Welsh et al. 2010).

Circadian rhythmicity is a cell-autonomous property. Rhythms are generated by a

molecular oscillator, consisting of clock genes and proteins that are organized in

positive and negative transcription translation feedback loops (TTFL) and cyclically

switch each other on and off (Lee et al. 2001). Briefly, in the positive limb of the TTFL,

a heterodimer of the basic loop-helix-loop/PAS (bHLH/PAS) domain containing

transcription factors circadian locomotor output cycles caput (CLOCK) and aryl

hydrocarbon receptor nuclear translocator-like (BMAL1) drives the transcription of

the Period (Per1 and Per2) and Cryptochrome (Cry1 and Cry2) genes. In the negative


Appendix 1

160

limb, the PER and CRY proteins shut down transcription of their own genes by

forming a heterodimeric complex that translocates to the nucleus and represses

CLOCK/BMAL1 driven transcription (Ko and Takahashi 2006; Lee et al. 2001; Reppert

and Weaver 2001). This molecular clock is coupled to output processes via a series

of clock-controlled genes, including transcription factors, cyclic expression of

which further relays rhythmicity to a wider set of genes (Kumaki et al. 2008; Miller

et al. 2007; Ueda et al. 2002; Yan et al. 2008). Indeed, various expression profiling

studies have shown that, depending on the tissue, up to 10% of the transcriptome

is under circadian control, with mRNA expression peaks distributed throughout

the circadian cycle (Akhtar et al. 2002; Bozek et al. 2009; Panda et al. 2002; Storch

et al. 2002).

The liver plays a key role in most metabolic processes, especially in detoxification

of xenobiotics. In this tissue, clock controlled transcription factors cause rhythmic

expression of members of the cytochrome P450 (CYP) family and other phase I

and phase II enzymes involved in this detoxification (Gachon et al. 2006). Adding

to circadian control over xenobiotic metabolism (both activation and inactivation)

is the diurnal variation in uptake and renal clearance of toxic substances (Levi and

Schibler 2007). Circadian variation in the absorption, metabolism and excretion

of toxic substances will ultimately determine the actual concentration to which

the cells and tissues are exposed (Paschos et al. 2010). As such, the time of day

of exposure to substances such as drugs, environmental chemicals and other

xenobiotics may affect the severity of the toxic response, a phenomenon referred

to as chronotoxicity.

To obtain insight into how the circadian clock affects the outcome of a

transcriptomics experiment, we have performed an in vivo study in which mice

were exposed to cyclophosphamide (CP) during the day or during the night. CP

is a chemotherapeutic and immune suppressive agent that is widely used for

treatment of several types of cancer, blood and bone marrow and tissue transplant

rejections, and autoimmune disorders (Colvin 1999). As a prodrug, CP requires

metabolic activation by cytochrome P450 enzymes (Hales 1982; Hill et al. 1972;

Ludeman 1999). The human cytochrome P450 enzymes CYP2B6, CYP2C9, and

CYP3A4/5 convert CP into the active metabolite 4-hydroxycyclophosphamide,

which is in tautomeric equilibrium with aldophosphamide. A small proportion of

aldophosphamide spontaneously decomposes to the genotoxic and carcinogenic



161

A1

metabolite phosphoramide mustard and the toxic byproduct acrolein (Hales 1982;

Kern and Kehrer 2002; King and Perry 2001). As shown here, analysis of the hepatic

gene expression profiles of mice exposed to CP during the day or during the

night revealed a remarkable time of day of exposure difference in the magnitude

of the response. These results highlight the involvement of the circadian clock in

genotoxic stress responses and indicate the importance to consider the circadian

system when performing toxicological risk analyses studies.

Methods

Animal experimentsThree-week-old male C57BL/6 mice were purchased from Harlan Laboratories

(Horst, The Netherlands) and housed at the animal facilities of the National

Institute for Public Health and the Environment (RIVM). Animals were kept under a

regular 12-hr light: 12-hr dark cycle and food and water were provided ad libitum.

All animal experiments were conducted in in accordance with national legislation

and were approved by a local ethical committee on experimental animals.

For the microarray study, 8-week-old mice received a single i.p. injection of either

300 mg/kg cyclophosphamide (CP, CAS 6055-19-2, Sigma-Aldrich) or the vehicle

(PBS) at two different Zeitgeber times, ZT8 and ZT20 (ZT0 = lights on). Animals

(n=4 per time point) were sacrificed by cervical dislocation at 0.5, 1, 2 and 4 hours

after CP injection (T0.5, T1, T2 and T4), and at 0 and 4 hours after vehicle injection

(C0 and C4). The left lobe of the liver was isolated and stored using RNAlater

(Invitrogen, Grand Island, NY, USA) according to the manufacturer’s protocol.

For the acute toxicity experiment, three dose groups were used: 30, 100 and 300

mg/kg CP. Eight-week-old mice (n=5 per group) received a single i.p. injection of

CP or the vehicle (PBS) at ZT8 or ZT20 and were sacrificed by orbital bleeding under

Ketamine/Xylazine anesthesia at 24 hours and 72 hours after exposure. Blood was

collected in both EDTA and gel collection tubes. Bone marrow was aspirated from

femur bones. The left lobe of the liver and the kidneys were removed and collected

in formalin.


Appendix 1

162

Microarray analysisRNA was extracted from all liver samples using the miRNeasy Mini Kit (Qiagen

Benelux, Venlo, The Netherlands). RNA concentrations were measured using a

NanoDrop ND-1000 Spectrophotometer (Nanodrop Technologies, Wilmington,

DE, USA), and RNA quality was assessed with an Agilent 2100 Bioanalyzer (Agilent

Technologies, Amstelveen, the Netherlands).

RNA was hybridized to Affymetrix HT Mouse Genome 430 PM Array plates

(Affymetrix Inc., Santa Clara, CA, USA) at the Microarray Department of the University

of Amsterdam, the Netherlands. In short, labeled cRNA samples were prepared

with the GeneChip 3” IVT express kit (Affymetrix) as described in the Affymetrix

GeneChip HT 3” IVT Express Technical Manual (Affymetrix), using 200 ng of purified

total RNA as template for the reaction. The array images were acquired using a

GeneChip HT Array Plate Scanner (Affymetrix) and analyzed with Affymetrix HT

software suite including expression console software (Affymetrix).

Data analysisThe raw data were subjected to a set of quality control checks. This quality check

excluded the presence of significant hybridization and experimental blocking

effects. All arrays passed quality control and were annotated according to de

Leeuw et al. (de Leeuw et al. 2008) and expression values were calculated using the

robust multi-array average (RMA) algorithm (Affy package, version 1.22.0; (Irizarry

et al. 2003)), available from the Bioconductor project (http://www.bioconductor.

org) for the R statistical language (http://cran.r-project.org). After normalization,

35,225 transcripts were used for Principal Component Analysis (PCA; produced in

R, version 2.13.1), including non-annotated transcripts and without cut-off for fold

change. All 35,225 transcripts were statistically analyzed for differential expression

using a mixed linear model with coefficients for each experimental group (Smyth

2004; Wolfinger et al. 2001). For each regime, a contrast analysis was applied to

compare each exposure time point with the vehicle control time point C0, and

to compare exposure time point T4 with the time-matched vehicle control C4.

For hypothesis testing, a permutation-based Fs test was used (Cui et al. 2005).

False discovery rate (FDR) correction per contrast was performed according to

Storey and Tibshirani (Storey and Tibshirani 2003), with an FDR<0.05 considered

as statistically significant. The gene expression results have been deposited in

the National Center for Biotechnology Information Gene Expression Omnibus

(Accession No. GSE46035).



163

A1

Enrichment of selected gene lists was analyzed with MetaCore Pathway Maps from

GeneGo, Inc. (http://www.genego.com/). Gene expression data were imported

into Metacore using their NCBI Entrez GeneID as identifier; therefore, only genes

annotated with a GeneID in the current NCBI database were used for pathway

enrichment analysis. Pathways were considered significantly enriched with an

FDR<0.01. Gene Set Enrichment Analysis allowed us to identify significant up- or

downregulation of enriched gene sets (P<0.05 and FDR<0.25) (Mootha et al. 2003;

Subramanian et al. 2005).

Histology, blood and bone marrow analysisParaffin-embedded, formalin fixed liver and kidney sections were stained

with Hematoxylin and Eosin (H&E) for histopathological evaluation. Aspartate

aminotransferase (AST) and alanine aminotransferase (ALT) levels were determined

with an auto-analyzer (Cobas Fara, Roche Diagnostics, Woerden, the Netherlands),

using dedicated kits from Roche Diagnostics. EDTA blood and bone marrow cells,

aspirated from femur bones with impulse cytophotometer solution (composed

as described previously in (Tonk et al. 2010)), were analyzed automatically using

the ADVIA 2120 (Siemens, Deerfield, USA). Bone marrow cytospin slides (Shandon

Cytospin 2, Thermo Fisher Scientific, Waltham, USA) were fixed with methanol,

stained according to May Grunwald and Giemsa and analyzed according to OECD

Test Guideline 474, Mammalian Erythrocyte Micronucleus Test (OECD). In short,

2000 polychromatic erythrocytes (PCEs) were scored per animal for the incidence

of micronucleated PCEs (MNPCEs). To determine myelotoxicity, 500 polychromatic

and normochromatic cells (NCE) were scored (PCE/NCE ratio).

All parameter values are represented as means ± SD, where appropriate. Statistical

analyses were performed using GraphPad Prism version 6.02 for Windows,

(GraphPad Software, La Jolla, California, USA). Comparisons between groups were

performed with one-way ANOVA or two-way ANOVA followed Dunnet’s test for

dose groups and Sidak’s test for differences between ZTs. P-values smaller than

0.05 were considered statistically significant.


Appendix 1

164

Results

Gene expression analysis - general impressionMale C57BL/6 mice, kept under a 12-hr light/12-hr dark cycle, were exposed to

cyclophosphamide (CP, 300 mg/kg body weight i.p.) at two different times during

the day, i.e. ZT8 and ZT20. Zeitgeber Time (ZT) is a standard experimental time

based on the period of the Zeitgeber, which under aforementioned light regime

equals a normal 24-hr day, and in which ZT0 represents “lights on” and ZT12 “lights

off”. At 0.5, 1, 2 and 4 hours after CP treatment (T0.5, T1, T2 and T4), and 0 and

4 hours after vehicle treatment (C0 and C4), animals were sacrificed and livers

were collected for gene expression analysis. To obtain a general impression of the

changes in transcription upon exposure to CP at different times of day, we first

visualized the overall response using Principal Component Analysis (PCA) (Figure

1). This analysis revealed two distinct clusters, reflecting exposure at the light (ZT8,

left half panel) and dark (ZT20, right half panel) phase, with the response becoming

stronger over time after exposure (C0, T0.5, T1, T2 and T4 top down). In addition,

differences in gene expression between the various time points appeared to be

more pronounced after exposure during the light phase (see Figure 1, left panel).

From these results, especially those for C4-T4, we conclude that time of day of

exposure has a significant impact on the overall response to CP exposure.

Analysis of differentially expressed genes and pathway enrichmentWe subsequently analyzed the gene expression data following a commonly used

toxicological approach, with ZT8-C0 and ZT20-C0 (time-zero) as the control time

points. Differences in hepatic gene expression between CP-treated animals and

the vehicle controls were calculated using a one-way ANOVA with an FDR<0.05.

This analysis resulted in an increasing number of differentially expressed genes

(DEGs) over time, with the highest number of DEGs observed 4 hours after

exposure: 5,837 and 2,942 DEGs (including non-annotated genes) at ZT8-T4 and

ZT20-T4, respectively (Figure 2).



165

A1

Figure 1. Principal Component Analysis using data of all genes. This analysis revealed two distinct clusters, representing the differential response of the genome to cyclophosphamide exposure at different times of day, ZT8 (circles) and ZT20 (triangles). The response was stronger over time after exposure (C0 to T4, top down); gene expression differences between various time points were more pronounced upon light phase exposure (left half panel). Open and closed symbols represent vehicle control and exposed samples, respectively.

Figure 2. Differentially expressed genes time-zero analysis. The number of differentially expressed genes, including non-annotated genes (DEGs; FDR<0.05) in the liver of CP-treated animals compared to those of vehicle controls obtained at C0. The number of DEGs increases over time, both for animals exposed during the light (ZT8, open bars) and the dark phase (ZT20, dashed bars). However, the number of significantly regulated genes is two- to four-fold lower upon exposure at ZT20 compared to animals exposed at ZT8.


Appendix 1

166

Despite identical exposure, two to four times as many DEGs were found when

animals were exposed at ZT8, as compared to exposure at ZT20. To investigate

whether the observed differences in the number of significantly regulated genes

also resulted in a different biological response, enrichment analyses of MetaCore

Pathways was performed. Gene Set Enrichment Analysis (GSEA) was used to obtain

an impression of up or down regulation of pathway groups. Since we observed

the highest number of significantly expressed genes 4 hours after exposure, we

focused on this time point. For these analyses, 4,089 and 2,084 DEGs (excluding

non-annotated genes) were used as input for ZT8 and ZT20, respectively. CP

exposure at ZT8 caused differential regulation of 125 pathways (Supplementary

Table S1a), most of which are consistent with the known effects of CP: genotoxicity,

oxidative stress, and immune modulation (Hussain et al. 2013; Rehman et al.

2012; Tripathi and Jena 2008). We found induction of the ATM/ATR pathway and

downregulation of pathways involved in cell cycle regulation, which are typically

associated with a DNA damage response invoked by the alkylating properties

of CP. Besides DNA damage, CP clearly induced oxidative stress as identified by

the overrepresentation and upregulation of intracellular signaling cascades such

as NF-ƙB, JAK/STAT, AKT/PI3K and MAPK pathways, all known to be activated

by oxidative stress. In addition, cell adhesion and cytoskeleton pathways were

enriched, probably in response to oxidative stress induced protein degradation.

The liver is predominantly an organ of innate immunity, favoring a general defense

response to stressors. Induction of several cytokine pathways, such as IL-1 and IL-6,

is indicative of this general immune or stress response. This response is most likely

coming from liver sinusoidal endothelial cells (LSEC) and Kupffer cells, which are

known to be more sensitive to CP than hepatocytes (DeLeve 1996).

Exposure at ZT20 resulted in a far lower number of significantly responding

pathways (i.e. 52 versus 125, see Supplementary Table S1b), which nonetheless

share a strong overall similarity to those observed upon exposure at ZT8.

Apparently, the response of the hepatic transcriptome is markedly weaker after

CP treatment at ZT20. Interestingly, a DNA damage response, clearly visible at ZT8,

was not significantly activated at ZT20 when using significantly regulated genes as

input (Supplementary Table S1b). The observed differences in pathway regulation

between the two time points may be due to the difference in number of genes

used as input for the enrichment analyses. We therefore matched the number of

input genes for ZT20 with ZT8, and used the top 4,089 genes (ranked according



167

A1

to FDR). This more balanced approach not only yielded an increased number of

pathways (from 52 up to 188), but, more importantly, also revealed a pattern of

responsive pathways that is similar to the one observed for ZT8 (Supplementary

Table S1c and S1a, respectively). Overall, for both time points, the DNA damage,

oxidative stress and immune response as described above could be identified

using this approach. Taken together, these pathway analyses suggest that the

response to CP exposure at ZT20 is comparable to the response observed at ZT8,

but that the effects in terms of gene expression changes are much stronger at ZT8.

Time-matched analysis of differentially expressed genes and pathway enrichmentAs evident from the PCA plot (Figure 1), the factor time of day of exposure resulted

in a distinct expression profile, since the profiles of the control group at time point

t=0 hours (C0) were markedly different from those at time point t=4 hours (C4),

both for ZT8 and ZT20. This indicates that also in non-exposed animals gene

expression patterns change over time, which is likely attributed to circadian clock

progression. Accordingly, the commonly time-zero analysis (as presented in the

previous section) identifies a mixture of (i) circadian clock controlled genes, (ii)

CP responsive genes, and (iii) expression of circadian genes influenced by CP

exposure. To take into account the circadian regulation of gene expression, we

repeated the aforementioned analyses with time-matched controls, 4 hours after

exposure (C4). As expected, this time-matched analysis resulted in a lower number

of DEGs for both ZT8 and ZT20, due to exclusion of genes that are under control

of the circadian clock and do not respond to CP. We cross-referenced DEGs for

time-zero analysis and time-matched analysis with clock controlled genes from 5

different literature data sets (Akhtar et al. 2002; Hughes et al. 2010; Miller et al. 2007;

Panda et al. 2002; Ueda et al. 2002). A loss of circadian regulated genes using the

time-matched analysis was observed. Remaining differentially expressed genes

are a mixture of CP responsive genes and circadian genes affected by CP exposure

The number of DEGS, however, remained higher at ZT8 than at ZT20, i.e. 2,015

versus 179 DEGs (including non-annotated genes) (Supplementary Figure S1).

To illustrate the effect of time-matched analysis in comparison to the time-

zero analysis, we examined the cytochrome P450 genes involved in the

biotransformation of CP (Cyp2b10, Cyp2c29, and Cyp3a13; the mouse homologues

of human CYP2B6, CYP2C9 and CYP3A4/5, respectively). Although, these genes


Appendix 1

168

Figure 3. Cytochrome P450 gene expression upon CP exposure. Expression of cytochrome P450 genes encoding enzymes involved in the metabolism of CP (Cyp2b10, Cyp2c29, and Cyp3a13; the mouse homologues of CYP2B6, CYP2C9 and CYP3A4/5, respectively) A. Fold Ratio (FR) of liver gene expression in CP-treated animals compared to that of vehicle control animals obtained at C0. All genes showed a time-dependent response to CP treatment from 0.5 until 4 hours after exposure. However, the magnitude of the response differed between light phase (ZT8) and dark phase exposure (ZT20). B. FR of liver gene expression



169

A1

of CP-treated animals 4 hours after exposure compared to time-matched vehicle controls obtained at C4. Only a marginal difference in Cyp2c29 gene expression induction seemed to be caused by time of day of exposure. Cyp2b10 and Cyp3a13 show no difference in gene expression between circadian phases upon CP exposure.

were significantly induced at both ZT8 and ZT20 when using time-zero controls,

fold ratios (induction) clearly differed between ZT8 and ZT20 exposure (Figure 3A).

However, time-matched analysis of the expression levels of these CP metabolizing

cytochrome P450 genes revealed that induction levels were comparable at both

circadian phases (Figure 3B). Other examples of genes for which a time-matched

analysis affects fold ratio differences between ZT8 and ZT20 exposed animals, in

comparison to time-zero analyses are shown in Supplementary Figure S2.

Analyses directed towards functional enrichment showed 79 significantly

responsive pathways for treatment at ZT8 (1,353 annotated genes, FDR<0.01 for

significant pathways; Supplementary Table S2a), of which 51 overlapped with

those found at this treatment time in the previous C0 control based pathway

analysis. The observed 79 pathways included pathways involved in DNA damage

response, as well as oxidative stress related pathways such as NF-ƙB and MAPK

signaling. However, cell adhesion and cytoskeleton remodeling pathways

associated with oxidative stress induced protein damage were not as pronounced

as compared to the analysis using C0 as a control. On the other hand, immune

related pathways showed a comparable induction as previously found. We did not

find any functional enrichment at ZT20, owing to the low number of input genes,

i.e. 112 annotated genes. Use of an equal number (top 1,353; ranked according

to FDR) of input genes resulted in only eight significantly enriched pathways

that, not surprisingly, only marginally reflect a response typical for CP exposure

(Supplementary Table S2c).

Thus, the use of time-matched controls uncovered marked differences in the

transcriptomic response of the liver when exposed at ZT20 and ZT8. Exposure at

ZT8 induced an unambiguous CP response, whereas the known CP effects could

hardly be detected after exposure at ZT20.

Time-zero versus time-matched analysis of pathway enrichmentThe pathway analyses described in the previous sections, using either time-zero

(C0) or time-matched (C4) controls, were performed with a different number of

input genes, which hampers a direct comparison of their effect. We therefore re-


Appendix 1

170

analyzed enrichment of responsive biological pathways for the time-zero data set,

using the same number of input genes as used for the time-matched analysis.

Using the top 1,353 annotated genes, we still noticed a strong CP mediated toxic

response after treatment at ZT8, as evident from the enrichment in DNA damage,

oxidative stress and immune responsive pathways, while such response now

appeared virtually absent when mice were treated at ZT20 (Supplementary Table

S3a and S3b, respectively). A comparison of the number of significantly enriched

MetaCore Pathways Maps, grouped per biological theme (Table 1) reveals that a

reduced number of input genes for the time zero analysis results in a differential

response at ZT8 and ZT20 that matches well with that observed when time-

matched controls are used. Taken together, our study, especially our time-matched

analysis, has shown that although CP exposure induces cytochrome P450 gene

expression equally at ZT8 and ZT20, a robust transcriptional response of the liver

is only visible at ZT8.

Table 1. Summary of MetaCore Pathway Maps enrichment

C0 controls C4 controls

ZT8 ZT20 ZT8 ZT20

Oxidative stress response 36 6 35 2

Immune response 17 1 20 0

DNA damage response 14 0 14 4

Other 2 0 0 2

Number of significantly enriched MetaCore Pathways Maps, grouped per biological

theme. Top 1,353 genes (ranked based on FDR) were used as input for this analysis.

Unless stated otherwise, pathways were grouped based on the main MetaCore

Pathway biological function groups (Supplementary Tables S1b and S1c).

Acute toxicity effectsTo correlate the observed changes in gene expression to more traditional

toxicological endpoints, we conducted an acute toxicity experiment with a similar

study design as used for the transcriptomics study. Animals were exposed to a

single dose of 0, 30, 100 or 300 mg/kg CP at ZT8 or ZT20 and sacrificed 24 and 72

hours after treatment. Body and organ weights were collected, and white blood

cell (WBC) counts and micronuclei formation were determined. In addition, we

performed a histopathological analysis of the liver and kidneys. Adverse effects of

CP became apparent by body weight loss at the highest doses of CP, irrespective



171

A1

of whether animals were treated at ZT8 or ZT20 (Supplementary Figure S3). We

also observed a slight, dose dependent decrease in kidney weight at 72 hours

after exposure in the ZT8 group, whereas no such effect was seen in the ZT20

group (Supplementary Figure S4). Histopathology of the kidney, however, did not

reveal any abnormalities in both groups. Likewise, we did not find any effect of

CP exposure on the liver for all parameters examined. Liver weight loss was not

observed (Supplementary Figure S5), and analysis of histopathology and serum

levels of the toxicity serum marker enzymes ALT and AST did not reveal any

abnormalities (data not shown).

However, CP exposure at ZT8 and at ZT20 resulted in a dose dependent decrease

in the overall number of white blood cells (Supplementary Table S4). This is largely

accounted for by a loss of lymphocytes, which showed a very similar response

(Figure 4). Exposure to a low dose of CP at ZT8 caused a significantly stronger

reduction in the number of lymphocytes after 24 hours than exposure at ZT20.

This effect however was no longer observed 72 hours after exposure. In bone

marrow, CP exposure resulted in a dose dependent induction of micronucleated

polychromatic erythrocytes (MNPCEs), a well-accepted parameter for genotoxicity,

except for the highest dose group (Supplementary Figure S6). Lack of micronuclei

formation at this dose is most likely due to myelotoxicity as indicated by the

decreased PCE/NCE ratio (Supplementary Figure S6). Despite the clear effect of

CP exposure on micronuclei formation, we did not find a difference in micronuclei

induction between the two ZT groups.


Appendix 1

172

Figure 4. Lymphocyte counts in peripheral blood of CP exposed mice. Lymphocyte counts were measured 24 and 72 hours after exposure. Counts are depicted relative to vehicle controls per circadian phase, values represent mean ± sd. Exposure to 30 mg/kg CP at ZT8 significantly more reduced the number of lymphocytes at 24 hours than in the ZT20 exposed animal (Sidak's multiple comparisons test, P<0.05).

Discussion

In toxicogenomic studies time of day of exposure of an animal to a (geno)toxic

chemical is usually not taken into account. However, it has become evidently clear

that the time of day of exposure can markedly influence the magnitude of the

biological response, which is referred to as chronotoxicity (Levi et al. 2010; Levi

and Schibler 2007), and which can be expected to translate into pronounced

differences in the outcome of transcriptomics experiments (Destici et al. 2009).

Likewise, with 10% of the active genes in a tissue being controlled by the clock,



173

A1

one can envisage that the choice of the control time point (i.e. a time-zero control,

taken at the start of exposure, or a time-matched control, taken when the animal

is sacrificed) can be of influence. In the current study, we addressed the impact of

the circadian clock on toxicogenomics experiments by exposing mice to a single

dose of cyclophosphamide (CP) at defined time-points during the day (ZT8) and

night (ZT20) and analyzing the response of the liver transcriptome.

Initially, following a commonly used toxicological approach that uses the zero

time points as a control, we observed a more than two-fold stronger response of

the liver transcriptome at ZT8 compared to ZT20, as evident from the number of

significantly differentially expressed genes (5,837 versus 2,942 DEGs). Subsequent

pathway analysis with the top 4,098 annotated genes revealed common CP

toxicity associated responses (i.e. DNA damage response, oxidative stress and

immune related pathways (Hussain et al. 2013; Rehman et al. 2012; Tripathi and

Jena 2008)), independent of the time of exposure. As expected, when repeating

the analysis using time-matched controls, we observed an overall reduction of the

number of differentially expressed genes at both ZT8 (from 5,837 to 2,015 DEGs)

and ZT20 (from 2,942 to 179 DEGs), partially because we filtered out the clock-

controlled, CP non-responsive DEGs. Strikingly, this approach uncovered a massive

impact of time of day effect of CP exposure on the hepatic transcriptome, being

10-fold more responsive at ZT8 than at ZT20 (2,015 versus 179 DEGs). Moreover,

after subjecting the top 1,353 annotated genes to pathway enrichment analysis,

the DNA damage, immune and oxidative stress responses, remained clearly

visible at ZT8, but were markedly less intense (DNA damage response), or virtually

absent (immune and oxidative stress response) at ZT20. Apparently, when time-

matched controls are omitted, circadian clock progression during the experiment

can partially blur the outcome of these experiments when DEGs are applied for

functional analysis. However, the identified responsive pathways using an equal

number of input genes as shown in Table 1 clearly show that the difference in liver

responsiveness to CP can be identified using C0 and C4 controls.

Our transcriptomics data point towards an increased responsiveness of the liver

to CP exclusively during the light phase. Upon exposure at ZT8, we observed a

marked induction of CP toxicity related pathways, whereas virtually no such

pathways were found for animals exposed at ZT20. Thus, the challenging question

arises which (clock-controlled) mechanisms are underlying this day and night


Appendix 1

174

difference in CP sensitivity. We have shown that CP exposure at ZT8 or ZT20 results

in a comparable upregulation of Cyp2b10, Cyp3a13 and Cyp2c29 expression, which

implies that the rate of CP metabolization does not depend on the time of drug

administration. Indeed, a pharmacokinetics study of Gorbacheva and coworkers

has shown that the formation and clearance of CP metabolites does not show

time of day dependency [38]. Accordingly, the presence and absence of a CP-

mediated toxic response at ZT8 and ZT20, respectively, is unlikely to originate from

differences in the induction of DNA damage and rather relates to a differential

magnitude of downstream events such as DNA damage signaling, DNA repair and

apoptosis.

It has been suggested that all aspects of the cellular response to DNA damage are

controlled or influenced by the circadian clock (Sancar et al. 2010). The CLOCK/

BMAL1 complex positively regulates the expression of the p21 gene, which plays

a critical role in the G1/S checkpoint (Grechez-Cassiau et al. 2008) as well as the

p53 gene, involved in DNA damage initiated or intrinsic apoptosis pathways

(Mullenders et al. 2009). Both p21 and p53 show low expression at the light dark

transition in the Mouse1.OST dataset of the CircaDB database (http://circadb.org)

(Pizarro et al. 2013). As Bmal1 gene expression is also at its lowest around ZT12, it

is tempting to speculate that this implies low expression of CLOCK/BMAL1 target

genes at ZT8, which in turn results in a greater dynamic range for the CP response.

This would fit with the observation that circadian clock deficient Cry1/Cry2 double

knockout mice (constitutive high expression of E-box genes) are more tolerant to

CP, while Clock mutant and Bmal1 knockout mice (constitutive low expression of

E-box genes) appear to be more sensitive (Gorbacheva et al. 2005).

Various animal studies have demonstrated that the time of administration can

cause changes in the biological outcome upon exposure to CP and some other

chemotherapeutic agents (Blumenthal et al. 2001; Gorbacheva et al. 2005; Granda

et al. 2001; Ohdo 2010). Compounds like celocoxib, doxorubicin and docetaxel were

best tolerated during the light phase, around ZT7 (Blumenthal et al. 2001; Granda

et al. 2001). Also for CP, wild type mice seemed to be more tolerant when exposed

at the end of the light phase (i.e. ZT10-ZT14) than when exposed at the end of the

dark phase (i.e. ZT18-ZT2), as demonstrated by a higher mortality rate (Gorbacheva

et al. 2005). Remarkably, our transcriptomics data contrast these survival data in

that the liver appears more sensitive to CP during the light phase. Upon exposure

at ZT8, we observed activation of CP toxicity related pathways, whereas virtually no



175

A1

such pathways were induced in the liver of animals exposed during the dark phase

(ZT20). This increase in sensitivity is supported by the slightly lower lymphocyte

counts in ZT8 exposed animals as compared to ZT20 exposed animals. In contrast,

the increased responsiveness of hepatic gene expression could be an enhanced

adaptive response to CP induced toxicity. As a result, ZT8 exposed animals are

better primed to CP exposure. Using gene-expression, we cannot differentiate

between increased sensitivity or enhanced adaptive response. Although we used

the same mouse strain, a major difference between our study and the study by

Gorbacheva and co-workers is the number of doses used. We employed a single

dose, whereas studies by Gorbacheva et al. were done using multiple doses of CP.

In addition, the time points used for the resting and the active phase were slightly

different: we used ZT8/ZT20 instead of ZT10-14/ZT22-2 as used by Gorbacheva,

respectively.

Despite the substantial changes in hepatic gene expression, we did not observe

pathological changes in the liver of CP exposed mice. CP exposure has been

reported to change the ultrastructural organization of the hepatocytes and SECs

(Lushnikova et al. 2011). Probably, a single dose is not able to induce pathological

changes or, alternatively, these changes only may become apparent at later

time points. C57Bl6 mice have been reported to be less sensitive to CP-induced

hepatoxicity than other mouse strains (Anton 1993). Other parameters, however,

such as body weight, kidneys weights, white blood cell counts (WBC) and

micronuclei induction were affected by CP exposure. CP is well-known for its toxic

effects, including immunosuppression and DNA damage induction (Emadi et al.

2009), as also uncovered by our transcriptomics study. For example, CP has been

reported to induce toxicity to B lymphocytes (Colvin 1999). Although in our study

lymphocytes were not differentiated, the decrease in WBC could be contributed

to the decrease in all lymphocytes. Overall, ZT8 exposed mice appeared to be

somewhat more sensitive than ZT20 animals. These findings indicate that gene

expression, in comparison to more traditional toxicity parameters, is more sensitive

to time of day effects. Possibly, the observed differences in transcription do become

apparent in apical end points at later time points.

In conclusion, this study highlights the importance of time of day in toxicity

testing, especially in toxicogenomics studies. To our knowledge, exposures in in

vivo rodent studies occur mainly during the day, which is comparable with ZT8


Appendix 1

176

exposure in our study. For risk-assessment purposes, it is of the utmost importance

to evaluate the worst-case scenario when estimating the risk of adverse health

effects induced by environmental chemicals. For assessment of CP toxicity in mice,

this seems to be exposure during the day. Whether ZT8 is the most sensitive time

point, and whether this sensitivity to toxic effects of chemicals is higher in mice

during the light phase in general, warrants further investigation. From a regulatory

point of view, it would also be worthwhile to examine possible effects of the

circadian clock on the outcome of in vitro toxicity tests. Furthermore, circadian

sensitivity to pharmaceuticals might also be beneficial when used in a clinical

setting, to minimize adverse effects and maximize efficacy, as has been shown for

certain anti-cancer drugs (Levi et al. 2011). Thus, considering the circadian clock in

toxicity testing might eventually lead to improved risk assessment and a more safe

use of chemical substances, as well as to improved medicine use.



177

A1

ReferencesAkhtar RA, Reddy AB, Maywood ES, et al. (2002) Circadian cycling of the mouse liver transcriptome, as

revealed by cDNA microarray, is driven by the suprachiasmatic nucleus. Current biology : CB 12(7):540-50 doi:S0960982202007595 [pii]

Anton E (1993) Differential sensitivity of DBA/2 and C57BL/6 mice to cyclophosphamide. Journal of applied toxicology : JAT 13(6):423-7

Balsalobre A, Damiola F, Schibler U (1998) A serum shock induces circadian gene expression in mammalian tissue culture cells. Cell 93(6):929-37 doi:S0092-8674(00)81199-X [pii]

Blumenthal RD, Waskewich C, Goldenberg DM, Lew W, Flefleh C, Burton J (2001) Chronotherapy and chronotoxicity of the cyclooxygenase-2 inhibitor, celecoxib, in athymic mice bearing human breast cancer xenografts. Clinical cancer research : an official journal of the American Association for Cancer Research 7(10):3178-85

Bozek K, Relogio A, Kielbasa SM, et al. (2009) Regulation of clock-controlled genes in mammals. PloS one 4(3):e4882 doi:10.1371/journal.pone.0004882

Colvin OM (1999) An overview of cyclophosphamide development and clinical applications. Current pharmaceutical design 5(8):555-60

Cui X, Hwang JT, Qiu J, Blades NJ, Churchill GA (2005) Improved statistical tests for differential gene expression by shrinking variance components estimates. Biostatistics 6(1):59-75 doi:6/1/59 [pii] 10.1093/biostatistics/kxh018

de Leeuw WC, Rauwerda H, Jonker MJ, Breit TM (2008) Salvaging Affymetrix probes after probe-level re-annotation. BMC Res Notes 1:66 doi:10.1186/1756-0500-1-66 1756-0500-1-66 [pii]

DeLeve LD (1996) Cellular target of cyclophosphamide toxicity in the murine liver: role of glutathione and site of metabolic activation. Hepatology 24(4):830-7 doi:10.1002/hep.510240414

Destici E, Oklejewicz M, Nijman R, Tamanini F, van der Horst GT (2009) Impact of the circadian clock on in vitro genotoxic risk assessment assays. Mutat Res 680(1-2):87-94 doi:10.1016/j.mrgentox.2009.09.001

Emadi A, Jones RJ, Brodsky RA (2009) Cyclophosphamide and cancer: golden anniversary. Nature reviews Clinical oncology 6(11):638-47 doi:10.1038/nrclinonc.2009.146

Gachon F, Olela FF, Schaad O, Descombes P, Schibler U (2006) The circadian PAR-domain basic leucine zipper transcription factors DBP, TEF, and HLF modulate basal and inducible xenobiotic detoxification. Cell metabolism 4(1):25-36 doi:10.1016/j.cmet.2006.04.015

Gorbacheva VY, Kondratov RV, Zhang R, et al. (2005) Circadian sensitivity to the chemotherapeutic agent cyclophosphamide depends on the functional status of the CLOCK/BMAL1 transactivation complex. Proc Natl Acad Sci U S A 102(9):3407-12 doi:0409897102 [pii] 10.1073/pnas.0409897102

Granda TG, Filipski E, D’Attino RM, et al. (2001) Experimental chronotherapy of mouse mammary adenocarcinoma MA13/C with docetaxel and doxorubicin as single agents and in combination. Cancer research 61(5):1996-2001

Grechez-Cassiau A, Rayet B, Guillaumond F, Teboul M, Delaunay F (2008) The circadian clock component BMAL1 is a critical regulator of p21WAF1/CIP1 expression and hepatocyte proliferation. J Biol Chem 283(8):4535-42 doi:10.1074/jbc.M705576200

Hales BF (1982) Comparison of the mutagenicity and teratogenicity of cyclophosphamide and its active metabolites, 4-hydroxycyclophosphamide, phosphoramide mustard, and acrolein. Cancer research 42(8):3016-21

Hill DL, Laster WR, Jr., Struck RF (1972) Enzymatic metabolism of cyclophosphamide and nicotine and production of a toxic cyclophosphamide metabolite. Cancer research 32(4):658-65

Hughes ME, Hogenesch JB, Kornacker K (2010) JTK_CYCLE: an efficient nonparametric algorithm for detecting rhythmic components in genome-scale data sets. Journal of biological rhythms 25(5):372-80 doi:10.1177/0748730410379711

Hussain A, Shadma W, Maksood A, Ansari SH (2013) Protective effects of Picrorhiza kurroa on cyclophosphamide-induced immunosuppression in mice. Pharmacognosy research 5(1):30-5 doi:10.4103/0974-8490.105646

Irizarry RA, Hobbs B, Collin F, et al. (2003) Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4(2):249-64 doi:10.1093/biostatistics/4.2.249 4/2/249 [pii]


Appendix 1

178

Kern JC, Kehrer JP (2002) Acrolein-induced cell death: a caspase-influenced decision between apoptosis and oncosis/necrosis. Chem Biol Interact 139(1):79-95 doi:S0009279701002952 [pii]

King PD, Perry MC (2001) Hepatotoxicity of chemotherapy. Oncologist 6(2):162-76

Ko CH, Takahashi JS (2006) Molecular components of the mammalian circadian clock. Hum Mol Genet 15 Spec No 2:R271-7 doi:15/suppl_2/R271 [pii] 10.1093/hmg/ddl207

Kumaki Y, Ukai-Tadenuma M, Uno KD, et al. (2008) Analysis and synthesis of high-amplitude Cis-elements in the mammalian circadian clock. Proc Natl Acad Sci U S A 105(39):14946-51 doi:10.1073/pnas.0802636105

Lee C, Etchegaray JP, Cagampang FR, Loudon AS, Reppert SM (2001) Posttranslational mechanisms regulate the mammalian circadian clock. Cell 107(7):855-67 doi:S0092-8674(01)00610-9 [pii]

Levi F, Karaboue A, Gorden L, et al. (2011) Cetuximab and circadian chronomodulated chemotherapy as salvage treatment for metastatic colorectal cancer (mCRC): safety, efficacy and improved secondary surgical resectability. Cancer chemotherapy and pharmacology 67(2):339-48 doi:10.1007/s00280-010-1327-8

Levi F, Okyar A, Dulong S, Innominato PF, Clairambault J (2010) Circadian timing in cancer treatments. Annual review of pharmacology and toxicology 50:377-421 doi:10.1146/annurev.pharmtox.48.113006.094626

Levi F, Schibler U (2007) Circadian rhythms: mechanisms and therapeutic implications. Annual review of pharmacology and toxicology 47:593-628 doi:10.1146/annurev.pharmtox.47.120505.105208

Lowrey PL, Takahashi JS (2011) Genetics of circadian rhythms in Mammalian model organisms. Advances in genetics 74:175-230 doi:10.1016/B978-0-12-387690-4.00006-4

Ludeman SM (1999) The chemistry of the metabolites of cyclophosphamide. Current pharmaceutical design 5(8):627-43

Lushnikova EL, Molodykh OP, Nepomnyashchikh LM, Bakulina AA, Sorokina YA (2011) Ultrastructurural picture of cyclophosphamide-induced damage to the liver. Bulletin of experimental biology and medicine 151(6):751-6

Miller BH, McDearmon EL, Panda S, et al. (2007) Circadian and CLOCK-controlled regulation of the mouse transcriptome and cell proliferation. Proc Natl Acad Sci U S A 104(9):3342-7 doi:0611724104 [pii]

10.1073/pnas.0611724104

Mohawk JA, Green CB, Takahashi JS (2012) Central and peripheral circadian clocks in mammals. Annual review of neuroscience 35:445-62 doi:10.1146/annurev-neuro-060909-153128

Mootha VK, Lindgren CM, Eriksson KF, et al. (2003) PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet 34(3):267-73 doi:10.1038/ng1180

Mullenders J, Fabius AW, Madiredjo M, Bernards R, Beijersbergen RL (2009) A large scale shRNA barcode screen identifies the circadian clock component ARNTL as putative regulator of the p53 tumor suppressor pathway. PloS one 4(3):e4798 doi:10.1371/journal.pone.0004798

OECD Test No. 474: Mammalian Erythrocyte Micronucleus Test. OECD Publishing

Ohdo S (2010) Chronopharmaceutics: pharmaceutics focused on biological rhythm. Biological & pharmaceutical bulletin 33(2):159-67

Panda S, Antoch MP, Miller BH, et al. (2002) Coordinated transcription of key pathways in the mouse by the circadian clock. Cell 109(3):307-20 doi:S0092867402007225 [pii]

Paschos GK, Baggs JE, Hogenesch JB, FitzGerald GA (2010) The role of clock genes in pharmacology. Annual review of pharmacology and toxicology 50:187-214 doi:10.1146/annurev.pharmtox.010909.105621

Pizarro A, Hayer K, Lahens NF, Hogenesch JB (2013) CircaDB: a database of mammalian circadian gene expression profiles. Nucleic Acids Res 41(Database issue):D1009-13 doi:10.1093/nar/gks1161

Rehman MU, Tahir M, Ali F, et al. (2012) Cyclophosphamide-induced nephrotoxicity, genotoxicity, and damage in kidney genomic DNA of Swiss albino mice: the protective effect of Ellagic acid. Mol Cell Biochem 365(1-2):119-27 doi:10.1007/s11010-012-1250-x

Reppert SM, Weaver DR (2001) Molecular analysis of mammalian circadian rhythms. Annu Rev Physiol 63:647-76 doi:10.1146/annurev.physiol.63.1.647 63/1/647 [pii]

Reppert SM, Weaver DR (2002) Coordination of circadian timing in mammals. Nature 418(6901):935-41 doi:10.1038/nature00965 nature00965 [pii]

Sancar A, Lindsey-Boltz LA, Kang TH, Reardon JT, Lee JH, Ozturk N (2010) Circadian clock control of the cellular response to DNA damage. FEBS letters 584(12):2618-25 doi:10.1016/j.febslet.2010.03.017



179

A1

Smyth GK (2004) Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3:Article3 doi:10.2202/1544-6115.1027

Storch KF, Lipan O, Leykin I, et al. (2002) Extensive and divergent circadian gene expression in liver and heart. Nature 417(6884):78-83 doi:10.1038/nature744 nature744 [pii]

Storey JD, Tibshirani R (2003) Statistical significance for genomewide studies. Proc Natl Acad Sci U S A 100(16):9440-5 doi:10.1073/pnas.1530509100

Subramanian A, Tamayo P, Mootha VK, et al. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102(43):15545-50 doi:10.1073/pnas.0506580102

Tonk EC, de Groot DM, Penninks AH, et al. (2010) Developmental immunotoxicity of methylmercury: the relative sensitivity of developmental and immune parameters. Toxicological sciences : an official journal of the Society of Toxicology 117(2):325-35 doi:10.1093/toxsci/kfq223

Tripathi DN, Jena GB (2008) Ebselen attenuates cyclophosphamide-induced oxidative stress and DNA damage in mice. Free radical research 42(11-12):966-77 doi:10.1080/10715760802566558

Ueda HR, Chen W, Adachi A, et al. (2002) A transcription factor response element for gene expression during circadian night. Nature 418(6897):534-9 doi:10.1038/nature00906

Weaver DR (1998) The suprachiasmatic nucleus: a 25-year retrospective. J Biol Rhythms 13(2):100-12

Welsh DK, Takahashi JS, Kay SA (2010) Suprachiasmatic nucleus: cell autonomy and network properties. Annu Rev Physiol 72:551-77 doi:10.1146/annurev-physiol-021909-135919

Wolfinger RD, Gibson G, Wolfinger ED, et al. (2001) Assessing gene significance from cDNA microarray expression data via mixed models. J Comput Biol 8(6):625-37 doi:10.1089/106652701753307520

Yan J, Wang H, Liu Y, Shao C (2008) Analysis of gene regulatory networks in the mammalian circadian rhythm. PLoS Comput Biol 4(10):e1000193 doi:10.1371/journal.pcbi.1000193


APPENDIX 2

Supplemental data


Appendix 2

182

Supplemental data Chapter 2

Figure S1. For the cross-sectional study (a) the differences in body weight increase between the groups was not significant (RM-ANOVA, group: F(1, 40) = 2.255, p=0.1410; time: F(17, 680) = 53.74, p<0.0001; interaction: F(17, 680) = 1.046, p=0.4049). The body weight gain is expressed relative to the body weights of animals at the start of the experiment (mean + SEM). (b) Food intake was slightly, but significantly (Two-way ANOVA, group: F(7, 108) = 0.4663, p<0.0001), lower in the animals kept under LD-inversion conditions (open bars) compared to the LD controls (closed bars) (*Sidak’s posttest p<0.05).


Supplemental data

183

A2

Figure S2. Left and right panels show sleep timing for LD controls at different days of the week and animals exposed to LD-inversions at different days after an LD-inversion, for light and dark phase, respectively. Day 0 corresponds to the day the light schedule inverses and day 1-6 indicate the number of days after the inversion. The timing of sleep continuously changes due to the alternating light schedule. Sleep in light instantly decreases at the day of the shift and subsequently increases the following days. Sleep in dark shows an inverse pattern. For separate analyses of the recorded weeks, see Figure S3a and Figure S3b.


Appendix 2

184


Supplemental data

185

A2Fig

ure

S3a.

Tot

al s

leep

and

sle

ep ti

min

g fo

r LD

con

trol

s at

diff

eren

t day

s of

the

wee

k.


Appendix 2

186


Supplemental data

187

A2

Fig

ure

S3b.

Tot

al s

leep

and

sle

ep ti

min

g fo

r ani

mal

s ex

pose

d to

wee

kly

LD-in

vers

ions

at s

ubse

quen

t day

s af

ter a

n LD

-inve

rsio

n.


Appendix 2

188

Figure S4. Hypotheses linking shift work with cancer (adapted from Fritschi et al., 2011). The present study underlines the importance of internal desynchronisation, sleep disruption and timing of food intake in the relationship between shift work and adverse health effects, especially breast cancer and obesity. Melatonin suppression and decreased vitamin D by less sunshine exposure were excluded in our model.

Table S1. Pathology data

Tumor spectra

LD LD-inversions

Total number of mice analyzed 20 21

Tumor-bearing mice 20 21

Mammary gland tumor-bearing mice 16 16

Mammary gland tumor# 18 17

Carcinoma 5 1

Fibrosarcoma/Carcinosarcoma 12 15

Intraepithelial neoplasia 1 1

Mammary gland hyperplasia - 2

Lymphosarcoma 3 5

Other tumors 5 7#Total number of mammary gland tumors per group, multiple tumors were found in some animals.


The supplemental data of chapter 3 comprise three tables that are too extensive

to be added in print. These tables are publically available via: http://onlinelibrary.

wiley.com/doi/10.1002/pmic.201100497/abstract


Supplemental data

189

A2


Supporting Information Figure S1. Schematic overview of the experimental set up and time of sample collection. White areas indicate lights on, grey areas indicate lights off.


Appendix 2

190

Supporting Information Figure S2. Venn-diagram of selected hepatic gene expression biomarkers using the three classifier algorithms. The consensus gene set consists of the 15 genes overlapping between the three algorithms.


Supplemental data

191

A2

Supporting Information Table S1. Results of prediction of the three classifier algorithms, using the consensus gene set (A-C). A summary of the prediction accuracy and other prediction parameters is shown in (D).

A. SVM

Non-CRD CRD

Experiment Non-CRD 23 1 24

CRD 0 24 24

23 25

B. PAM-R

Non-CRD CRD


CRD 0 24 24

21 27

C. RF

Non-CRD CRD


CRD 2 22 24

23 25

D. SVM PAM-R RF

Accuracy 97.9% 93.6% 89.6%

Sensitivity 100% 100% 91.7%

Specificity 95.8% 87.5% 87.5%

Positive predictive value 96.0% 88.9% 88.0%

Negative predictive value 100% 100% 91.3%

Supporting Information Table S2. Ability of the consensus gene set to correctly classify phase-shifted animals as CRD after one shift, six shifts or five days recovery using the three classifier algorithms.

Sensitivity

SVM PAM-R RF

1 shift 33.3% 29.2% 29.2%

6 shifts 87.5% 91.7% 83.3%

5 days recovery 83.3% 87.5% 75.0%

Supporting Information Table S3. Sequential approach gene selection.

Supplemental Table S3 is too extensive to be added in print. This table is publically

available via: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4436131/


Appendix 2

192


Supplementary Table S1. Participant characteristics. Age is provided in years (y), weight in kilograms (kg), waist circumference in centimeters (cm), total sleep duration between the first sampling time at 08:00 AM and the last sampling time at 04:00 AM in hours (hr), sleep timing of sleep episodes between the first sampling time at 08:00 AM and the last sampling time at 04:00 AM provided in clock time (CT).

Participant Gender Age (y) Weight (kg)

Waist circumference (cm)

Total sleep duration (hr) between 08:00 and 04:00

Sleep timing (CT) between 08:00 and 04:00

F1 f 22 85 80 1.25 02:00-03:15

F2 f 22 52 68 5.42 13:30-14:30, 22:00-23:30, 00:45-3:40

F3 f 22 50 83 3.00 00:15-03:15

F4 f 22 49 69 3.00 00:15-03:15

F5 f 22 48 67 3.00 00:15-03:15

F6 f 22 50 69 3.00 00:15-03:15

F7 f 21 59 75 6.50 09:00-10:00, 17:00-19:00, 00:15-03:45

F8 f 21 55 77 4.75 16:15-18:00, 00:15-03:15

F9 f 21 82 93 0.00 none

F10 f 22 55 74 2.50 00:30-03:00

M1 m 41 105 123 4.00 14:00:15:30, 00:30-03:00

M2 m 24 100 110 3.00 00:30-03:30

M3 m 23 72 86 2.00 01:00-03:00

M4 m 21 70 83 5.00 20:30-23:30, 01:00-03:00

M5 m 21 88 85 3.00 00:15-03:15

M6 m 23 79 79 3.33 00:30-03:50

M7 m 21 75 90 3.67 00:10-03:50

Females

Mean 21.70 58.50 75.50 3.24

SD 0.48 13.61 8.17 1.91

Males

Mean 24.86 84.14 93.71 3.43

SD 7.22 13.90 16.35 0.94


Supplemental data

193

A2

Supplementary Table S2. Overview of methods used to determine parameters in plasma and serum. The intra assay variation (CV) was determined with three quality control samples (N=8) for the markers as determined on the auto-analyzer and with two quality control samples (N=5) for markers as measured on the immune-analyzer. The CVs of markers measured with the Luminex technique were obtained from the manufacturer.

Technique CV Source

ACTH Luminex (Millipore) 10.8 Plasma

CORT Immune-analyzer (Access-2, Beckman-Coulter) 6.2 Serum

DHEAS Immune-analyzer (Access-2, Beckman-Coulter) 3.9 Serum

E2 Immune-analyzer (Access-2, Beckman-Coulter) 6.3 Serum

FSH Luminex (Millipore) 7.2 Plasma

hGH Immune-analyzer (Access-2, Beckman-Coulter) 2.3 Serum

LH Luminex (Millipore) 6.3 Plasma

PRL Immune-analyzer (Access-2, Beckman-Coulter) 1.8 Serum

PRG Immune-analyzer (Access-2, Beckman-Coulter) 10.8 Serum

TEST Immune-analyzer (Access-2, Beckman-Coulter) 1.4 Serum

TotT3 Immune-analyzer (Access-2, Beckman-Coulter) 4.2 Serum

TSH Immune-analyzer (Access-2, Beckman-Coulter) 1.4 Serum

FFA Auto-analyzer (Unicel DxC 800, Beckman-Coulter) 4.4 Plasma

HDL Auto-analyzer (Unicel DxC 800, Beckman-Coulter) 4.0 Plasma

LDL Auto-analyzer (Unicel DxC 800, Beckman-Coulter) 7.2 Plasma

TG Auto-analyzer (Unicel DxC 800, Beckman-Coulter) 4.3 Plasma

CHOL Auto-analyzer (Unicel DxC 800, Beckman-Coulter) 4.5 Plasma


Appendix 2

194

Supplementary Table S3. Overview of hormonal parameters in serum or plasma for A. CircWave analysis of circadian rhythms; B. Repeated-Measures ANOVA to determine daily variation. Amplitude is presented as the % of the median. Ct = Clock time.

A. Circwave analysis

males females

Markers p-value peak (CT) amplitude p-value peak (CT) amplitude

ACTH 0.219 17:54 6.12% 0.714 17:22 3.71%

CORT 0.000 07:47 129.11% 0.000 07:29 161.63%

DHEAS 0.861 20:13 21.89% 0.571 18:23 21.36%

E2 0.306 06:51 24.66% 0.685 05:52 25.11%

FSH 0.964 02:03 40.47% 0.741 18:41 31.20%

hGH 0.895 04:13 581.23% 0.953 11:31 397.11%

LH 0.391 00:36 59.22% 0.968 07:15 45.25%

PRL 0.194 04:55 49.11% 0.081 06:20 97.09%

PRG 0.003 07:52 117.78% 0.870 06:17 44.70%

TEST 0.055 07:33 38.38% 0.282 07:38 22.75%

TotT3 0.025 06:18 10.14% 0.041 07:04 8.08%

TSH 0.001 01:54 75.89% 0.006 03:22 74.26%

B. RM-ANOVA all time points

males females

Markers p-value F-value p-value F-value

ACTH 0.288 F (2.277, 13.66) = 1.375 0.215 F (3.037, 27.33) = 1.775

CORT 0.003 F (2.611, 15.67) = 7.614 0.000 F (3.466, 31.19) = 20.56

DHEAS 0.045 F (2.389, 14.33) = 3.694 0.002 F (2.943, 26.49) = 6.530

E2 0.438 F (2.760, 16.56) = 0.938 0.154 F (3.873, 34.85) = 1.792

FSH 0.517 F (1.990, 11.94) = 0.695 0.512 F (2.079, 18.71) = 0.687

hGH 0.184 F (1.465, 8.792) = 2.091 0.459 F (2.243, 20.18) = 0.838

LH 0.198 F (2.794, 16.76) = 1.746 0.061 F (3.491, 31.42) = 2.617

PRL 0.014 F (2.715, 16.29) = 4.984 0.118 F (1.865, 16.79) = 2.461

PRG 0.020 F (2.310, 13.86) = 4.974 0.272 F (2.204, 19.84) = 1.395

TEST 0.021 F (2.945, 17.67) = 4.221 0.023 F (2.760, 24.84) = 3.922

TotT3 0.032 F (2.461, 14.77) = 4.078 0.078 F (2.896, 26.06) = 2.570

TSH 0.037 F (1.266, 7.596) = 5.991 0.008 F (1.700, 15.30) = 7.305


Supplemental data

195

A2

Supplementary Table S4. Overview of lipid parameters in serum or plasma for A. CircWave analysis of circadian rhythms; B. Repeated-Measures ANOVA to determine daily variation. Amplitude is presented as the % of the median. Ct = Clock time.

A. Circwave analysis

males females


FFA 0.012 08:31 74.91% 0.36 06:56 80.70%

HDL 0.695 07:38 9.20% 0.97 07:30 3.86%

LDL 0.818 08:07 12.02% 0.96 08:46 8.35%

TG 0.105 13:46 70.36% 0.05 14:13 50.33%

CHOL 0.645 08:47 11.44% 0.83 08:57 8.17%

B. RM-ANOVA all time points

males females

Markers p-value F-value p-value F-value

FFA 0.069 F (2.442, 14.65) = 3.078 0.221 F (2.638, 23.75) = 1.589

HDL 0.110 F (2.708, 16.25) = 2.399 0.213 F (2.475, 22.28) = 1.641

LDL 0.002 F (2.875, 17.25) = 7.746 0.006 F (2.660, 23.94) = 5.670

TG 0.048 F (1.943, 11.66) = 4.012 0.063 F (2.846, 25.62) = 2.798

CHOL 0.003 F (2.381, 14.29) = 8.618 0.003 F (2.938, 26.44) = 6.020

Supplementary Table S5. CircWave analysis of circadian rhythms classical circadian markers in serum (CORT) or PBMCs (PER1, BMAL1). Amplitude is presented as the % of the median. Ct = Clock time.

Circwave analysis

males females


CORT 0.000 07:47 129.11% 0.000 07:29 161.63%

PER1 0.002 08:07 80.15 %

BMAL1 0.035 14:12 35.55 %

Supporting information S6. Raw data files of data presented in this study.

Supporting information S6 is too extensive to be added in print. This table is

publically available via: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4540433/


Appendix 2

196


Supplemental Table S1. Genes oscillating under both LD and CRD conditions

Symbol

Gm13889

Ypel2

Chka

Nedd4l

Asap2

Dbp

Dnmt3b

Mif4gd

1110067D22Rik

Arntl

Clpx

Rorc

Per3

St5

A530001N23Rik

Fgfr4

Wdr91

Larp1

Dnajb11

Gramd4

Slc45a3

Tenc1

Hsd17b7

Nr1d2

D7Wsu130e

Prkd3

Gm129

9830001H06Rik

Pdcd2l

Stx2

Lingo4

Dynll2

Hist1h4i


Supplemental data

197

A2

Supplemental Table S2. Significantly enriched processes based on 561 genes that lost significant circadian rhythmicity after CRD exposure.

Processes Groupentrainment of circadian clock by photoperiod Circadian rhythm

entrainment of circadian clock Circadian rhythm

photoperiodism Circadian rhythm

circadian regulation of gene expression Circadian rhythm

regulation of circadian rhythm Circadian rhythm

circadian rhythm Circadian rhythm

rhythmic process Circadian rhythm

epithelial cell proliferation involved in mammary gland duct elongation Hormonal regulation

Sertoli cell proliferation Hormonal regulation

mammary gland branching involved in pregnancy Hormonal regulation

branch elongation involved in mammary gland duct branching Hormonal regulation

negative regulation of gonadotropin secretion Hormonal regulation

vagina development Hormonal regulation

branching involved in prostate gland morphogenesis Hormonal regulation

cellular response to follicle-stimulating hormone stimulus Hormonal regulation

Sertoli cell development Hormonal regulation

response to follicle-stimulating hormone Hormonal regulation

androgen metabolic process Hormonal regulation

cellular response to estradiol stimulus Hormonal regulation

Sertoli cell differentiation Hormonal regulation

cellular response to gonadotropin stimulus Hormonal regulation

cellular hormone metabolic process Hormonal regulation

regulation of endocrine process Hormonal regulation

hormone-mediated signaling pathway Hormonal regulation

response to testosterone Hormonal regulation

negative regulation of peptide hormone secretion Hormonal regulation

ovarian follicle development Hormonal regulation

ovulation cycle process Hormonal regulation

negative regulation of hormone secretion Hormonal regulation

ovulation cycle Hormonal regulation

female gonad development Hormonal regulation

hormone metabolic process Hormonal regulation

response to estradiol Hormonal regulation

female sex differentiation Hormonal regulation

development of primary female sexual characteristics Hormonal regulation

regulation of peptide hormone secretion Hormonal regulation

regulation of hormone levels Hormonal regulation

male sex differentiation Hormonal regulation

steroid metabolic process Hormonal regulation

cellular response to hormone stimulus Hormonal regulation

cellular response to peptide hormone stimulus Hormonal regulation

regulation of hormone secretion Hormonal regulation

response to estrogen Hormonal regulation

response to peptide hormone Hormonal regulation

response to steroid hormone Hormonal regulation


Appendix 2

198

Processes Group

negative regulation of mitosis Cell cycle

negative regulation of nuclear division Cell cycle

DNA damage response, signal transduction by p53 class mediator Cell cycle

transforming growth factor beta receptor signaling pathway Cell cycle

signal transduction in response to DNA damage Cell cycle

regulation of mitosis Cell cycle

cellular response to transforming growth factor beta stimulus Cell cycle

regulation of nuclear division Cell cycle

response to transforming growth factor beta Cell cycle

negative regulation of cell cycle process Cell cycle

regulation of cell division Cell cycle

negative regulation of cell cycle Cell cycle

regulation of mitotic cell cycle Cell cycle

regulation of cell cycle process Cell cycle

mitotic cell cycle process Cell cycle

mitotic cell cycle Cell cycle

negative regulation of cell proliferation Cell cycle

lipid storage Glucose and lipid metabolism

positive regulation of phospholipase activity Glucose and lipid metabolism

regulation of glucose import Glucose and lipid metabolism

positive regulation of glucose import Glucose and lipid metabolism

regulation of glycogen metabolic process Glucose and lipid metabolism

positive regulation of lipase activity Glucose and lipid metabolism

regulation of phospholipase activity Glucose and lipid metabolism

positive regulation of glucose transport Glucose and lipid metabolism

negative regulation of insulin secretion Glucose and lipid metabolism

regulation of glucose transport Glucose and lipid metabolism

regulation of lipase activity Glucose and lipid metabolism

monosaccharide metabolic process Glucose and lipid metabolism

regulation of insulin secretion Glucose and lipid metabolism

glucose metabolic process Glucose and lipid metabolism

hexose metabolic process Glucose and lipid metabolism

regulation of lipid metabolic process Glucose and lipid metabolism

lipid biosynthetic process Glucose and lipid metabolism

coumarin metabolic process Resp. to exogenous compound

phenylpropanoid metabolic process Resp. to exogenous compound

flavonoid glucuronidation Resp. to exogenous compound

flavonoid biosynthetic process Resp. to exogenous compound

flavonoid metabolic process Resp. to exogenous compound

cellular glucuronidation Resp. to exogenous compound

glucuronate metabolic process Resp. to exogenous compound

uronic acid metabolic process Resp. to exogenous compound


Supplemental data

199

A2

Processes Group

quinone metabolic process Resp. to exogenous compound

catechol-containing compound biosynthetic process Resp. to exogenous compound

catecholamine biosynthetic process Resp. to exogenous compound

secondary metabolic process Resp. to exogenous compound

response to dexamethasone Resp. to exogenous compound

water-soluble vitamin metabolic process Resp. to exogenous compound

xenobiotic metabolic process Resp. to exogenous compound

cellular response to xenobiotic stimulus Resp. to exogenous compound

vitamin metabolic process Resp. to exogenous compound

response to xenobiotic stimulus Resp. to exogenous compound

isoprenoid metabolic process Resp. to exogenous compound

response to purine-containing compound Resp. to exogenous compound

organic hydroxy compound biosynthetic process Resp. to exogenous compound

response to ethanol Resp. to exogenous compound

monocarboxylic acid metabolic process Resp. to exogenous compound

response to alcohol Resp. to exogenous compound

response to nutrient Resp. to exogenous compound

response to drug Resp. to exogenous compound

cellular response to nitrogen compound Resp. to exogenous compound

ephrin receptor signaling pathway Various signaling pathways

positive regulation of epidermal growth factor receptor signaling pathway

Various signaling pathways

positive regulation of ERBB signaling pathway Various signaling pathways

ERK1 and ERK2 cascade Various signaling pathways

cellular response to growth hormone stimulus Various signaling pathways

cellular response to cAMP Various signaling pathways

response to cAMP Various signaling pathways

response to organophosphorus Various signaling pathways

transmembrane receptor protein serine/threonine kinase signaling pathway

Various signaling pathways

regulation of stress-activated MAPK cascade Various signaling pathways

regulation of stress-activated protein kinase signaling cascade Various signaling pathways

activation of MAPK activity Various signaling pathways

regulation of ERK1 and ERK2 cascade Various signaling pathways

regulation of JNK cascade Various signaling pathways

cellular response to fibroblast growth factor stimulus Various signaling pathways

neurotrophin TRK receptor signaling pathway Various signaling pathways

response to glucocorticoid Various signaling pathways

response to fibroblast growth factor Various signaling pathways

MAPK cascade Various signaling pathways

neurotrophin signaling pathway Various signaling pathways

response to corticosteroid Various signaling pathways


Appendix 2

200

Processes Group

cellular response to peptide Various signaling pathways

response to hypoxia Various signaling pathways

response to decreased oxygen levels Various signaling pathways

transmembrane receptor protein tyrosine kinase signaling pathway Various signaling pathways

response to oxygen levels Various signaling pathways

response to metal ion Various signaling pathways

positive regulation of MAPK cascade Various signaling pathways

cellular response to organonitrogen compound Various signaling pathways

cellular response to growth factor stimulus Various signaling pathways

response to peptide Various signaling pathways

regulation of MAPK cascade Various signaling pathways

response to growth factor Various signaling pathways

response to extracellular stimulus Various signaling pathways

positive regulation of intracellular signal transduction Various signaling pathways

iron ion transmembrane transport Other

positive regulation of myelination Other

CDP-diacylglycerol biosynthetic process Other

regulation of oligodendrocyte progenitor proliferation Other

CDP-diacylglycerol metabolic process Other

positive regulation of neurological system process Other

regulation of transmission of nerve impulse Other

negative regulation of heart rate Other

positive regulation of bone remodeling Other

positive regulation of bone resorption Other

regulation of smooth muscle cell apoptotic process Other

response to lead ion Other

regulation of calcium ion import Other

negative regulation of smooth muscle cell proliferation Other

positive regulation of phospholipase C activity Other

regulation of phospholipase C activity Other

negative regulation of muscle cell apoptotic process Other

negative regulation of heart contraction Other

positive regulation of heart contraction Other

regulation of muscle cell apoptotic process Other

osteoclast differentiation Other

transferrin transport Other

positive regulation of multicellular organism growth Other

positive regulation of fibroblast proliferation Other

regulation of bone resorption Other

ferric iron transport Other

trivalent inorganic cation transport Other


Supplemental data

201

A2

Processes Group

cellular aldehyde metabolic process Other

regulation of smooth muscle cell proliferation Other

regulation of phosphatidylinositol 3-kinase activity Other

activation of phospholipase C activity Other

regulation of bone remodeling Other

protein targeting to mitochondrion Other

cellular ketone metabolic process Other

establishment of protein localization to mitochondrion Other

iron ion transport Other

pigment metabolic process Other

positive regulation of blood pressure Other

negative regulation of peptide secretion Other

protein localization to mitochondrion Other

regulation of neurological system process Other

regulation of fibroblast proliferation Other

actin filament organization Other

nucleoside monophosphate biosynthetic process Other

coenzyme biosynthetic process Other

regulation of cellular amine metabolic process Other

regulation of heart rate Other

transition metal ion transport Other

mitochondrial transport Other

cofactor biosynthetic process Other

tissue remodeling Other

exocytosis Other

actin cytoskeleton organization Other

actin filament-based process Other

cofactor metabolic process Other

cell-cell junction organization Other

regulation of peptide secretion Other

regulation of peptide transport Other

phospholipid biosynthetic process Other

coenzyme metabolic process Other

activation of protein kinase activity Other

positive regulation of ion transport Other

organophosphate biosynthetic process Other

regulation of neuron apoptotic process Other

regulation of blood pressure Other

regulation of actin cytoskeleton organization Other

regulation of actin filament-based process Other

regulation of homeostatic process Other


Appendix 2

202

Processes Group

regulation of neuron death Other

phospholipid metabolic process Other

positive regulation of kinase activity Other

positive regulation of transferase activity Other

positive regulation of protein kinase activity Other

secretion by cell Other

regulation of cytoskeleton organization Other

small molecule biosynthetic process Other

organonitrogen compound biosynthetic process Other

cellular metal ion homeostasis Other

secretion Other

metal ion homeostasis Other

cellular cation homeostasis Other

positive regulation of protein phosphorylation Other

cation homeostasis Other

cellular ion homeostasis Other

regulation of protein serine/threonine kinase activity Other

regulation of organelle organization Other

negative regulation of cellular component organization Other

regulation of kinase activity Other

ion homeostasis Other

regulation of protein kinase activity Other

positive regulation of multicellular organismal process Other

positive regulation of monocyte chemotaxis Other

response to interleukin-15 Other

acute-phase response Other

acute inflammatory response Other

platelet degranulation Other


Supplemental data

203

A2

Supplementary Table S3. Top 20 genes with largest decrease in amplitude.

Symbol Diff. amplitude

Upp2 -0.5235

Sybu -0.4985

Atxn7l1 -0.4623

Pdk4 -0.4521

Slc16a5 -0.4486

Ccrn4l -0.4224

Nr1d1 -0.4044

Etnk2 -0.3957

Igfbp1 -0.388

Stbd1 -0.3809

Wfdc2 -0.3807

Serpina6 -0.3761

Aqp4 -0.3759

B130006D01Rik -0.3631

Olfr1134 -0.362

Ppard -0.3608

Cyp2s1 -0.3605

Wdr81 -0.3541

Mcm10 -0.3462

Itsn1 -0.3454

Supplementary Table S4. Genes differentially expressed in the CRD group compared to LD controls.

Symbol p-value FC

Nespas 2.96E-06 -0.1539

Gm128 6.37E-05 -0.3601

Rfx7 9.32E-05 0.1652

Defa-rs1 0.000136 0.1718

Adamts5 0.000168 -0.1777

Rhox4e 0.000192 -0.1678

Sprr1a 0.000235 -0.2050

Gm1123 0.000316 -0.1293

Mpped2 0.00036 0.1585

9630013A20Rik 0.00043 0.1606

Gm10295 0.00052 -0.1364

LOC100503435 0.000674 0.1059

1190028D05Rik 0.000778 0.1334

Olfr1248 0.000815 0.1640

Mixl1 0.000868 -0.1246

Glt25d2 0.000912 -0.2461

Gm10773 0.000912 0.1299


Appendix 2

204

Supplemental Table S5. Overrepresentation of up or down regulated genes in GO processes or the C2 gene sets.

Upregulated

REACTOME_MEIOSIS Cell cycle

REACTOME_ACTIVATION_OF_ATR_IN_RESPONSE_TO_REPLICATION_STRESS DNA damage

KEGG_MISMATCH_REPAIR DNA damage

REACTOME_G2_M_CHECKPOINTS DNA damage

PID_FANCONI_PATHWAY DNA damage

REACTOME_DNA_STRAND_ELONGATION DNA replication

KEGG_DNA_REPLICATION DNA replication

REACTOME_EXTENSION_OF_TELOMERES DNA replication

REACTOME_LAGGING_STRAND_SYNTHESIS DNA replication

DNA_DEPENDENT_ATPASE_ACTIVITY DNA replication

REACTOME_MEIOTIC_RECOMBINATION DNA replication

REACTOME_TELOMERE_MAINTENANCE DNA replication

REACTOME_ACTIVATION_OF_THE_PRE_REPLICATIVE_COMPLEX DNA replication

REACTOME_CHROMOSOME_MAINTENANCE DNA replication

REACTOME_INFLUENZA_VIRAL_RNA_TRANSCRIPTION_AND_REPLICATION Immune

REACTOME_INFLUENZA_LIFE_CYCLE Immune

GLUTATHIONE_TRANSFERASE_ACTIVITY Metabolism

REACTOME_PHASE_II_CONJUGATION Metabolism

REACTOME_GLUTATHIONE_CONJUGATION Metabolism

KEGG_GLUTATHIONE_METABOLISM Metabolism

REACTOME_CHOLESTEROL_BIOSYNTHESIS Metabolism

REACTOME_METABOLISM_OF_PROTEINS Metabolism

REACTOME_NONSENSE_MEDIATED_DECAY_ENHANCED_BY_THE_EXON_JUNCTION_COMPLEX

Transcription

REACTOME_METABOLISM_OF_RNA Transcription

REACTOME_METABOLISM_OF_MRNA Transcription

REACTOME_CLEAVAGE_OF_GROWING_TRANSCRIPT_IN_THE_TERMINATION_REGION_

Transcription

KEGG_RIBOSOME Translation

REACTOME_SRP_DEPENDENT_COTRANSLATIONAL_PROTEIN_TARGETING_TO_MEMBRANE

Translation

REACTOME_PEPTIDE_CHAIN_ELONGATION Translation

STRUCTURAL_CONSTITUENT_OF_RIBOSOME Translation

REACTOME_TRANSLATION Translation

REACTOME_3_UTR_MEDIATED_TRANSLATIONAL_REGULATION Translation

TRANSFERASE_ACTIVITY_TRANSFERRING_ALKYL_OR_ARYLOTHER_THAN_METHYLGROUPS

Various

KEGG_ALZHEIMERS_DISEASE Various

STRUCTURAL_MOLECULE_ACTIVITY Various

REGULATION_OF_MUSCLE_CONTRACTION Various


Supplemental data

205

A2

Downregulated

REACTOME_APOPTOSIS Cell cycle

REACTOME_REGULATION_OF_MITOTIC_CELL_CYCLE Cell cycle

REACTOME_SCF_BETA_TRCP_MEDIATED_DEGRADATION_OF_EMI1 Cell cycle

REACTOME_AUTODEGRADATION_OF_THE_E3_UBIQUITIN_LIGASE_COP1 Cell cycle

REACTOME_APC_C_CDC20_MEDIATED_DEGRADATION_OF_MITOTIC_PROTEINS Cell cycle

REACTOME_ADAPTIVE_IMMUNE_SYSTEM Immune

REACTOME_INTERFERON_GAMMA_SIGNALING Immune

REACTOME_CYTOKINE_SIGNALING_IN_IMMUNE_SYSTEM Immune

REACTOME_INTERFERON_SIGNALING Immune

REACTOME_INTERFERON_ALPHA_BETA_SIGNALING Immune

REACTOME_SIGNALING_BY_THE_B_CELL_RECEPTOR_BCR Immune

KEGG_ACUTE_MYELOID_LEUKEMIA Immune

REACTOME_ACTIVATION_OF_NF_KAPPAB_IN_B_CELLS Immune

KEGG_CHEMOKINE_SIGNALING_PATHWAY Immune

KEGG_PRIMARY_IMMUNODEFICIENCY Immune

REACTOME_ANTIGEN_ACTIVATES_B_CELL_RECEPTOR_LEADING_TO_GENERATION_OF_SECOND_MESSENGERS

Immune

REACTOME_CROSS_PRESENTATION_OF_SOLUBLE_EXOGENOUS_ANTIGENS_ENDOSOMES

Immune

BIOCARTA_IL2_PATHWAY Immune

REACTOME_IMMUNOREGULATORY_INTERACTIONS_BETWEEN_A_LYMPHOID_AND_A_NON_LYMPHOID_CELL

Immune

KEGG_B_CELL_RECEPTOR_SIGNALING_PATHWAY Immune

REACTOME_CLASS_I_MHC_MEDIATED_ANTIGEN_PROCESSING_PRESENTATION Immune

REACTOME_IL_2_SIGNALING Immune

PID_TCR_PATHWAY Immune

DEFENSE_RESPONSE Immune

KEGG_T_CELL_RECEPTOR_SIGNALING_PATHWAY Immune

BIOCARTA_NFAT_PATHWAY Immune

REACTOME_DOWNSTREAM_SIGNALING_EVENTS_OF_B_CELL_RECEPTOR_BCR Immune

REACTOME_ER_PHAGOSOME_PATHWAY Immune

REACTOME_INTEGRIN_ALPHAIIB_BETA3_SIGNALING Immune

PID_IL4_2PATHWAY Immune

REACTOME_ANTIGEN_PROCESSING_UBIQUITINATION_PROTEASOME_DEGRADATION

Immune

PID_INTEGRIN_A4B1_PATHWAY Immune

REACTOME_CYTOCHROME_P450_ARRANGED_BY_SUBSTRATE_TYPE Metabolism

PROTEIN_CATABOLIC_PROCESS Metabolism

CELLULAR_PROTEIN_CATABOLIC_PROCESS Metabolism

MONOOXYGENASE_ACTIVITY Metabolism

BIOPOLYMER_CATABOLIC_PROCESS Metabolism

REACTOME_SYNTHESIS_OF_BILE_ACIDS_AND_BILE_SALTS Metabolism

PID_ERBB1_INTERNALIZATION_PATHWAY Proliferation

PID_MET_PATHWAY Proliferation

PID_PDGFRBPATHWAY Proliferation


Appendix 2

206

Downregulated Continued

TRANSMEMBRANE_RECEPTOR_PROTEIN_SERINE_THREONINE_KINASE_SIGNALING_PATHWAY

Proliferation

REACTOME_SIGNALING_BY_WNT Proliferation

REACTOME_SIGNALING_BY_SCF_KIT Proliferation

PID_AURORA_A_PATHWAY Proliferation

KEGG_ERBB_SIGNALING_PATHWAY Proliferation

PID_ECADHERIN_KERATINOCYTE_PATHWAY Proliferation

KEGG_PROTEASOME Proteasome

PID_PI3KCIPATHWAY Proteasome

REACTOME_REGULATION_OF_ORNITHINE_DECARBOXYLASE_ODC Proteasome

AMINE_TRANSPORT Transport

AMINO_ACID_TRANSPORT Transport

INTRACELLULAR_TRANSPORT Transport

CYTOPLASMIC_VESICLE_PART Transport

CYTOPLASMIC_VESICLE_MEMBRANE Transport

AMINE_TRANSMEMBRANE_TRANSPORTER_ACTIVITY Transport

RNA_BINDING Various

PID_TCPTP_PATHWAY Various

MITOCHONDRION Various

MEMBRANE_FRACTION Various

CYTOSKELETON Various

ENZYME_LINKED_RECEPTOR_PROTEIN_SIGNALING_PATHWAY Various

PROTEIN_TRANSPORT Transport

BIOCARTA_TPO_PATHWAY Various

POST_TRANSLATIONAL_PROTEIN_MODIFICATION Various

REACTOME_METABOLISM_OF_AMINO_ACIDS_AND_DERIVATIVES Various

REACTOME_SIGNAL_TRANSDUCTION_BY_L1 Various

PHOSPHORYLATION Various

KEGG_FOCAL_ADHESION Various

MITOCHONDRIAL_PART Various

ACTIN_POLYMERIZATION_AND_OR_DEPOLYMERIZATION Various

REGULATION_OF_PROTEIN_METABOLIC_PROCESS Various

CELL_FRACTION Various

MICROTUBULE Various

PROTEIN_TARGETING Various


working around the clock - repub.eur.nl · working around the clock: adverse health effects of...

Documents